<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>安全学术圈</title>
    <link>https://wechat2rss.xlab.app/feed/8c5d5f0004e7231abeb01dac49cac5da4ec6933d.xml</link>
    <description>分享安全方面的论文写作、会议发表、基金申请方面的资料。&#xA;(wechat feed made by @ttttmr https://wechat2rss.xlab.app)</description>
    <managingEditor> (安全学术圈)</managingEditor>
    <image>
      <url>https://wx.qlogo.cn/mmhead/Q3auHgzwzM5QdYXnIyaJT9H4VVHz2DKKQYEU5B9ZalqRslFPN3kCmw/0</url>
      <title>安全学术圈</title>
      <link>https://wechat2rss.xlab.app/feed/8c5d5f0004e7231abeb01dac49cac5da4ec6933d.xml</link>
    </image>
    <item>
      <title>中国科学院信息工程研究所 | SafeHarness：告别防御盲区，为智能体Harness穿上「全生命周期安全护甲」</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495331&amp;idx=1&amp;sn=d17385aa84b1614fd8e41d291ec61f2c</link>
      <description>SafeHarness 框架。这是一种将防御机制直接编织进智能体 Harness 运行生命周期的安全架构，它不仅能从根本上解决现有安全工具的结构性缺陷，还能在面对复合攻击时实现系统级的协同防御。</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-05-11 16:21</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=2d138973&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnw2bHYk6NlicpQJ59eEY3uRHeibT2ssfibmlrxxfvhz4mLoH7nJBwg5lcquiaRJV0kxE7QxDluibPkHoFwZ0aQRREw1MHVWaCgJz2xw%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>SafeHarness 框架。这是一种将防御机制直接编织进智能体 Harness 运行生命周期的安全架构，它不仅能从根本上解决现有安全工具的结构性缺陷，还能在面对复合攻击时实现系统级的协同防御。</p>
  <div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0px 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;overflow-wrap: break-word;text-align: justify;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#34;PingFang SC&#34;, Cambria, Cochin, Georgia, Times, &#34;Times New Roman&#34;, serif;" data-pm-slice="0 0 []"><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.30925925925925923" data-type="png" data-w="1080" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011681" src="https://wechat2rss.xlab.app/img-proxy/?k=a49667fb&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnwjA6Tk4uFlB8S64nWFwWqlQC5DPtnvBcFaLPa0ZQnbfQnErHBovW8zghibibWzDOxArepSnTqt8nTJS7vqzKOIOwx5F9IiaMF4ics%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><blockquote style="border: none;display: block;font-size: 0.9em;overflow: auto;overflow-scrolling: touch;border-left: 3px solid rgba(0, 0, 0, 0.4);background: rgba(0, 0, 0, 0.05);color: #6a737d;padding-top: 10px;padding-bottom: 10px;padding-left: 20px;padding-right: 10px;margin-bottom: 20px;margin-top: 20px;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;color: black;line-height: 26px;"><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">原文标题：SAFEHARNESS: Lifecycle-Integrated Security Architecture for LLM-based Agent Deployment</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">原文作者：Xixun Lin, Yang Liu, Yancheng Chen, Yongxuan Wu, Yucheng Ning, Yilong Liu, Nan Sun, Shun Zhang, Bin Chong, Chuan Zhou, Yanan Cao</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">原文链接：<a href="https://arxiv.org/abs/2604.13630" target="_blank">https://arxiv.org/abs/2604.13630</a></span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">笔记作者：宁钰成@中国科学院信息工程研究所</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">主编：黄诚@安全学术圈</span></span></em></p></blockquote><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">1. 背景介绍</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如果说大语言模型（LLM）是智能体的“大脑”，那么 Harness 就是支撑其在现实世界运转的躯干与骨骼。它绝不仅仅是一个简单的调用外壳，而是作为“指挥中心”，一手包办了工具编排、上下文管理和状态持久化等核心任务。然而，正是这种“大权在握”的核心地位，使得 Harness 成为了极具价值的攻击面：哪怕只是框架层面的单点妥协（例如一条被投毒的工具输出），风险也会顺着执行管道级联放大到整个系统。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">针对这一严峻挑战，</span><strong style="font-weight: bold;color: black;"><span leaf="">中国科学院信息工程研究所联合多家研究单位</span></strong><span leaf="">正式推出了 </span><strong style="font-weight: bold;color: black;"><span leaf="">SafeHarness</span></strong><span leaf=""> 框架。这是一种将防御机制直接编织进智能体 Harness 运行生命周期的安全架构，它不仅能从根本上解决现有安全工具的结构性缺陷，还能在面对复合攻击时实现系统级的协同防御。</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">2. 现有 Agent 安全防御的三大“致命伤”</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">当前的智能体安全工具（如外部的安全护栏）往往与智能体 Harness 存在严重的结构性错位。它们大多只是套在模型外层的“壳”，面临三大局限：</span></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">上下文盲区（Context Blindness）：</span></strong><span leaf=""> 现有的防御机制运行在框架边界之外。它们只在对话接口处过滤输入和输出，根本看不见框架内部的状态。这就像安检员只检查旅客的随身物品，却不知道旅客在候机厅里接触过什么危险源。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">层间孤岛（Inter-layer Isolation）：</span></strong><span leaf=""> 即使部署了多重安全检查，它们也是各自为战。输入过滤器发现了可疑内容，并不会通知下游的动作验证器提高警惕；被拦截的危险动作，也不会提醒记忆系统去怀疑触发该动作的上下文。面对复合攻击，孤立的检查点如同盲人摸象。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">缺乏韧性（Lack of Resilience）：</span></strong><span leaf=""> 现有的防御大多是一锤子买卖的“通过/拦截”。一旦攻击突破了外层防御，系统缺乏渐进式限制工具权限或优雅降级的机制，错误会随着执行步骤不断累积。</span></p></li></ol><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">3. SafeHarness：四层防御，步步为营</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">为了解决上述痛点，SafeHarness 提出了一个核心洞察：智能体执行的四个阶段（输入处理、决策制定、动作执行、状态更新）有着各自独特的安全需求，必须由专属的、感知阶段的防御层来守护。</span></p><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" alt="SafeHarness 架构图，展示了环绕在智能体周围的 L1 至 L4 防御层，以及跨层的信息反馈机制。" class="rich_pages wxw-img" data-ratio="0.6018518518518519" data-type="png" data-w="1080" style="display: block;margin: 0px auto;max-width: 100%;" data-imgfileid="100011672" src="https://wechat2rss.xlab.app/img-proxy/?k=3075efb5&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnxMlAcob7HXbPXg7aVSkZezPXXeAyDzsW1ibpdL83HDaH6tl8SSE57ce2biad6BVkOyXlcn9Y7ylLmroXX0Ykz7QjUP5Sia79FFlY%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span><figcaption style="margin-top: 5px;text-align: justify;color: rgb(136, 136, 136);font-size: 14px;"><span leaf="">SafeHarness 架构图，展示了环绕在智能体周围的 L1 至 L4 防御层，以及跨层的信息反馈机制。</span></figcaption></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">SafeHarness 打造了紧密耦合的四大防护层：</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">L1: INFORM（输入处理层）：</span></strong><span leaf=""> 在外部内容进入推理链前进行全面清洗。该层包含了</span><strong style="font-weight: bold;color: black;"><span leaf="">结构化净化</span></strong><span leaf="">（去除隐藏的Unicode字符防止逃逸） 、</span><strong style="font-weight: bold;color: black;"><span leaf="">模式匹配检测</span></strong><span leaf="">（快速拦截已知的注入指令） 以及</span><strong style="font-weight: bold;color: black;"><span leaf="">语义过滤</span></strong><span leaf="">（通过LLM辅助提取事实、剥离恶意指令）。同时，它会为每个数据块打上</span><strong style="font-weight: bold;color: black;"><span leaf="">溯源标签（Provenance Tagging）</span></strong><span leaf="">，让下游层清楚地知道这些数据是来自用户输入、工具输出还是检索内容，并动态评估信任度。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">L2: VERIFY（决策制定层）：</span></strong><span leaf=""> 对每一个工具调用提议应用</span><strong style="font-weight: bold;color: black;"><span leaf="">三级递进式验证级联</span></strong><span leaf="">。Tier 1 使用基于规则的引擎进行快速合规检查；Tier 2 调用“上下文裁判模型”进行多步分析；如果裁判感到模棱两可（UNCERTAIN），则升级到 Tier 3 进行深度的</span><strong style="font-weight: bold;color: black;"><span leaf="">因果诊断（Causal Diagnostics）</span></strong><span leaf="">，判定这个动作究竟是源于用户的真实意图，还是被恶意注入所误导。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">L3: CONSTRAIN（动作执行层）：</span></strong><span leaf=""> 在执行边界强制落实最小权限原则。工具被划分为五个风险层级；系统引入了受限的 </span><strong style="font-weight: bold;color: black;"><span leaf="">能力令牌（Capability Tokens）</span></strong><span leaf="">机制，带有生存时间（TTL）和调用次数限制。此外，还会执行基于 </span><strong style="font-weight: bold;color: black;"><span leaf="">HMAC-SHA256 的完整性校验</span></strong><span leaf="">，确保工具描述未被黑客在注册表中篡改。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">L4: CORRECT（状态更新层）：</span></strong><span leaf=""> 提供最后的防线。它会定期捕获环境和内存的</span><strong style="font-weight: bold;color: black;"><span leaf="">状态检查点</span></strong><span leaf="">。一旦 VERIFY 层确诊了攻击，CORRECT 会立即触发</span><strong style="font-weight: bold;color: black;"><span leaf="">安全回滚</span></strong><span leaf="">。最精妙的是其 </span><strong style="font-weight: bold;color: black;"><span leaf="">自适应降级（Adaptive Degradation）</span></strong><span leaf="">机制：回滚后系统不会直接崩溃，而是降低一个运行层级（例如禁用破坏性工具），并在连续安全运行一段时间后自动恢复权限，在安全与可用性之间取得完美平衡。</span></p></li></ul><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">4. 灵魂组件：跨层机制与层间反馈</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">四层防御不再是各自为战。SafeHarness 设计了全局视角的跨层机制（Cross-Layer Mechanisms） ，并通过</span><strong style="font-weight: bold;color: black;"><span leaf="">层间反馈（Inter-Layer Feedback）</span></strong><span leaf="">将整个系统连成一个有机的“免疫网络” ：</span></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">两大跨层基石：</span></strong></p></li><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">溯源保护内存：</span></strong><span leaf="">所有的对话历史、中间结果都被存储在只能追加（append-only）、带有内容哈希的内存中，彻底防止内存被恶意覆盖和伪造。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">熵监视器（Entropy Monitor）：</span></strong><span leaf="">这是一个滑动窗口，实时追踪系统近期的“无序度”（违规动作的比率）。</span></p></li></ul><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">动态层间反馈链路：</span></strong></p></li><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">L1 发现，L2 戒备（</span></strong><span style="cursor: pointer;"><span role="presentation" data-formula=" L1\rightarrow L2 " data-formula-type="inline-equation" style=""><svg xmlns="http://www.w3.org/2000/svg" role="img" focusable="false" viewBox="0 -683 3917.6 694" aria-hidden="true" style="vertical-align: -0.025ex;width: 8.863ex;height: 1.57ex;"><g stroke="currentColor" fill="currentColor" stroke-width="0" transform="matrix(1 0 0 -1 0 0)"><g data-mml-node="math"><g data-mml-node="mi"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(681, 0)"><path data-c="31" d="M213 578L200 573Q186 568 160 563T102 556H83V602H102Q149 604 189 617T245 641T273 663Q275 666 285 666Q294 666 302 660V361L303 61Q310 54 315 52T339 48T401 46H427V0H416Q395 3 257 3Q121 3 100 0H88V46H114Q136 46 152 46T177 47T193 50T201 52T207 57T213 61V578Z"></path></g><g data-mml-node="mo" transform="translate(1458.8, 0)"><path data-c="2192" d="M56 237T56 250T70 270H835Q719 357 692 493Q692 494 692 496T691 499Q691 511 708 511H711Q720 511 723 510T729 506T732 497T735 481T743 456Q765 389 816 336T935 261Q944 258 944 250Q944 244 939 241T915 231T877 212Q836 186 806 152T761 85T740 35T732 4Q730 -6 727 -8T711 -11Q691 -11 691 0Q691 7 696 25Q728 151 835 230H70Q56 237 56 250Z"></path></g><g data-mml-node="mi" transform="translate(2736.6, 0)"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(3417.6, 0)"><path data-c="32" d="M109 429Q82 429 66 447T50 491Q50 562 103 614T235 666Q326 666 387 610T449 465Q449 422 429 383T381 315T301 241Q265 210 201 149L142 93L218 92Q375 92 385 97Q392 99 409 186V189H449V186Q448 183 436 95T421 3V0H50V19V31Q50 38 56 46T86 81Q115 113 136 137Q145 147 170 174T204 211T233 244T261 278T284 308T305 340T320 369T333 401T340 431T343 464Q343 527 309 573T212 619Q179 619 154 602T119 569T109 550Q109 549 114 549Q132 549 151 535T170 489Q170 464 154 447T109 429Z"></path></g></g></g></svg></span></span><strong style="font-weight: bold;color: black;"><span leaf="">）：</span></strong><span leaf="">当 L1 探测到注入企图或判定内容不可信时，不仅会拦截，还会推高系统“熵值”，使得后续 L2 层直接绕过简单的规则检查，强制启动更严格的 LLM 级核查。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">L2 确诊，L4 抢救（</span></strong><span style="cursor: pointer;"><span role="presentation" data-formula=" L2\rightarrow L4 " data-formula-type="inline-equation" style=""><svg xmlns="http://www.w3.org/2000/svg" role="img" focusable="false" viewBox="0 -683 3917.6 694" aria-hidden="true" style="vertical-align: -0.025ex;width: 8.863ex;height: 1.57ex;"><g stroke="currentColor" fill="currentColor" stroke-width="0" transform="matrix(1 0 0 -1 0 0)"><g data-mml-node="math"><g data-mml-node="mi"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(681, 0)"><path data-c="32" d="M109 429Q82 429 66 447T50 491Q50 562 103 614T235 666Q326 666 387 610T449 465Q449 422 429 383T381 315T301 241Q265 210 201 149L142 93L218 92Q375 92 385 97Q392 99 409 186V189H449V186Q448 183 436 95T421 3V0H50V19V31Q50 38 56 46T86 81Q115 113 136 137Q145 147 170 174T204 211T233 244T261 278T284 308T305 340T320 369T333 401T340 431T343 464Q343 527 309 573T212 619Q179 619 154 602T119 569T109 550Q109 549 114 549Q132 549 151 535T170 489Q170 464 154 447T109 429Z"></path></g><g data-mml-node="mo" transform="translate(1458.8, 0)"><path data-c="2192" d="M56 237T56 250T70 270H835Q719 357 692 493Q692 494 692 496T691 499Q691 511 708 511H711Q720 511 723 510T729 506T732 497T735 481T743 456Q765 389 816 336T935 261Q944 258 944 250Q944 244 939 241T915 231T877 212Q836 186 806 152T761 85T740 35T732 4Q730 -6 727 -8T711 -11Q691 -11 691 0Q691 7 696 25Q728 151 835 230H70Q56 237 56 250Z"></path></g><g data-mml-node="mi" transform="translate(2736.6, 0)"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(3417.6, 0)"><path data-c="34" d="M462 0Q444 3 333 3Q217 3 199 0H190V46H221Q241 46 248 46T265 48T279 53T286 61Q287 63 287 115V165H28V211L179 442Q332 674 334 675Q336 677 355 677H373L379 671V211H471V165H379V114Q379 73 379 66T385 54Q393 47 442 46H471V0H462ZM293 211V545L74 212L183 211H293Z"></path></g></g></g></svg></span></span><strong style="font-weight: bold;color: black;"><span leaf="">）：</span></strong><span leaf="">如果 L2 因果诊断确诊了攻击，L4 就会立刻执行时空回滚并降低系统特权等级。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">L4 降级，L3 锁死（</span></strong><span style="cursor: pointer;"><span role="presentation" data-formula=" L4\rightarrow L3 " data-formula-type="inline-equation" style=""><svg xmlns="http://www.w3.org/2000/svg" role="img" focusable="false" viewBox="0 -683 3917.6 705" aria-hidden="true" style="vertical-align: -0.05ex;width: 8.863ex;height: 1.595ex;"><g stroke="currentColor" fill="currentColor" stroke-width="0" transform="matrix(1 0 0 -1 0 0)"><g data-mml-node="math"><g data-mml-node="mi"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(681, 0)"><path data-c="34" d="M462 0Q444 3 333 3Q217 3 199 0H190V46H221Q241 46 248 46T265 48T279 53T286 61Q287 63 287 115V165H28V211L179 442Q332 674 334 675Q336 677 355 677H373L379 671V211H471V165H379V114Q379 73 379 66T385 54Q393 47 442 46H471V0H462ZM293 211V545L74 212L183 211H293Z"></path></g><g data-mml-node="mo" transform="translate(1458.8, 0)"><path data-c="2192" d="M56 237T56 250T70 270H835Q719 357 692 493Q692 494 692 496T691 499Q691 511 708 511H711Q720 511 723 510T729 506T732 497T735 481T743 456Q765 389 816 336T935 261Q944 258 944 250Q944 244 939 241T915 231T877 212Q836 186 806 152T761 85T740 35T732 4Q730 -6 727 -8T711 -11Q691 -11 691 0Q691 7 696 25Q728 151 835 230H70Q56 237 56 250Z"></path></g><g data-mml-node="mi" transform="translate(2736.6, 0)"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(3417.6, 0)"><path data-c="33" d="M127 463Q100 463 85 480T69 524Q69 579 117 622T233 665Q268 665 277 664Q351 652 390 611T430 522Q430 470 396 421T302 350L299 348Q299 347 308 345T337 336T375 315Q457 262 457 175Q457 96 395 37T238 -22Q158 -22 100 21T42 130Q42 158 60 175T105 193Q133 193 151 175T169 130Q169 119 166 110T159 94T148 82T136 74T126 70T118 67L114 66Q165 21 238 21Q293 21 321 74Q338 107 338 175V195Q338 290 274 322Q259 328 213 329L171 330L168 332Q166 335 166 348Q166 366 174 366Q202 366 232 371Q266 376 294 413T322 525V533Q322 590 287 612Q265 626 240 626Q208 626 181 615T143 592T132 580H135Q138 579 143 578T153 573T165 566T175 555T183 540T186 520Q186 498 172 481T127 463Z"></path></g></g></g></svg></span></span><strong style="font-weight: bold;color: black;"><span leaf="">）：</span></strong><span leaf="">L4 下达的降级指令，会直接决定 L3 允许使用的工具上限，高危工具瞬间被物理隔离。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">安全期，自动复苏（</span></strong><span style="cursor: pointer;"><span role="presentation" data-formula=" L4\rightarrow L4/L2 " data-formula-type="inline-equation" style=""><svg xmlns="http://www.w3.org/2000/svg" role="img" focusable="false" viewBox="0 -750 5598.6 1000" aria-hidden="true" style="vertical-align: -0.566ex;width: 12.666ex;height: 2.262ex;"><g stroke="currentColor" fill="currentColor" stroke-width="0" transform="matrix(1 0 0 -1 0 0)"><g data-mml-node="math"><g data-mml-node="mi"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(681, 0)"><path data-c="34" d="M462 0Q444 3 333 3Q217 3 199 0H190V46H221Q241 46 248 46T265 48T279 53T286 61Q287 63 287 115V165H28V211L179 442Q332 674 334 675Q336 677 355 677H373L379 671V211H471V165H379V114Q379 73 379 66T385 54Q393 47 442 46H471V0H462ZM293 211V545L74 212L183 211H293Z"></path></g><g data-mml-node="mo" transform="translate(1458.8, 0)"><path data-c="2192" d="M56 237T56 250T70 270H835Q719 357 692 493Q692 494 692 496T691 499Q691 511 708 511H711Q720 511 723 510T729 506T732 497T735 481T743 456Q765 389 816 336T935 261Q944 258 944 250Q944 244 939 241T915 231T877 212Q836 186 806 152T761 85T740 35T732 4Q730 -6 727 -8T711 -11Q691 -11 691 0Q691 7 696 25Q728 151 835 230H70Q56 237 56 250Z"></path></g><g data-mml-node="mi" transform="translate(2736.6, 0)"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(3417.6, 0)"><path data-c="34" d="M462 0Q444 3 333 3Q217 3 199 0H190V46H221Q241 46 248 46T265 48T279 53T286 61Q287 63 287 115V165H28V211L179 442Q332 674 334 675Q336 677 355 677H373L379 671V211H471V165H379V114Q379 73 379 66T385 54Q393 47 442 46H471V0H462ZM293 211V545L74 212L183 211H293Z"></path></g><g data-mml-node="TeXAtom" data-mjx-texclass="ORD" transform="translate(3917.6, 0)"><g data-mml-node="mo"><path data-c="2F" d="M423 750Q432 750 438 744T444 730Q444 725 271 248T92 -240Q85 -250 75 -250Q68 -250 62 -245T56 -231Q56 -221 230 257T407 740Q411 750 423 750Z"></path></g></g><g data-mml-node="mi" transform="translate(4417.6, 0)"><path data-c="4C" d="M228 637Q194 637 192 641Q191 643 191 649Q191 673 202 682Q204 683 217 683Q271 680 344 680Q485 680 506 683H518Q524 677 524 674T522 656Q517 641 513 637H475Q406 636 394 628Q387 624 380 600T313 336Q297 271 279 198T252 88L243 52Q243 48 252 48T311 46H328Q360 46 379 47T428 54T478 72T522 106T564 161Q580 191 594 228T611 270Q616 273 628 273H641Q647 264 647 262T627 203T583 83T557 9Q555 4 553 3T537 0T494 -1Q483 -1 418 -1T294 0H116Q32 0 32 10Q32 17 34 24Q39 43 44 45Q48 46 59 46H65Q92 46 125 49Q139 52 144 61Q147 65 216 339T285 628Q285 635 228 637Z"></path></g><g data-mml-node="mn" transform="translate(5098.6, 0)"><path data-c="32" d="M109 429Q82 429 66 447T50 491Q50 562 103 614T235 666Q326 666 387 610T449 465Q449 422 429 383T381 315T301 241Q265 210 201 149L142 93L218 92Q375 92 385 97Q392 99 409 186V189H449V186Q448 183 436 95T421 3V0H50V19V31Q50 38 56 46T86 81Q115 113 136 137Q145 147 170 174T204 211T233 244T261 278T284 308T305 340T320 369T333 401T340 431T343 464Q343 527 309 573T212 619Q179 619 154 602T119 569T109 550Q109 549 114 549Q132 549 151 535T170 489Q170 464 154 447T109 429Z"></path></g></g></g></svg></span></span><strong style="font-weight: bold;color: black;"><span leaf="">）：</span></strong><span leaf="">警报解除、连续动作安全后，系统会自动调高权限，并同步放松 L2 的审查力度，恢复常态运行。</span></p></li></ul></ol><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">5. 硬核战绩：安全与任务效用的完美平衡</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">研究团队在包含 2,000 个安全敏感任务的 Agent-SafetyBench 上，测试了三种主流智能体框架（ReAct, Multi-Agent, Self-Evolving）和四种安全基线，并引入了涵盖六大威胁类别的五种攻击场景。</span></p><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" alt="不同架构下各安全模式的主实验结果对比，展示了 DeepSeek-Chat 与 DeepSeek-Reasoner 在无保护、Guardrail、LlamaFirewall 及 SafeHarness 等不同模式下的 UBR、ASR、TCR 等核心指标。" class="rich_pages wxw-img" data-ratio="0.3592592592592593" data-type="png" data-w="1080" style="display: block;margin: 0px auto;max-width: 100%;" data-imgfileid="100011675" src="https://wechat2rss.xlab.app/img-proxy/?k=e767963c&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnzcxjVqxcdNUgIjyg8V8nuqzRHibwS8iat2slSPO7lJpavaY64YzZ0YuHYCAeXjJEyNNExPBX64XMyJQOZeyasDtQ0f7icDpdIibMg%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span><figcaption style="margin-top: 5px;text-align: justify;color: rgb(136, 136, 136);font-size: 14px;"><span leaf="">不同架构下各安全模式的主实验结果对比，展示了 DeepSeek-Chat 与 DeepSeek-Reasoner 在无保护、Guardrail、LlamaFirewall 及 SafeHarness 等不同模式下的 UBR、ASR、TCR 等核心指标。</span></figcaption></figure><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" alt="不同框架与安全模式下的不安全行为率 (UBR) 对比条形图。在 DeepSeek-Chat 和 DeepSeek-Reasoner 模型下，SafeHarness 在所有框架中的不安全行为率均为最低。" class="rich_pages wxw-img" data-ratio="0.3824074074074074" data-type="png" data-w="1080" style="display: block;margin: 0px auto;max-width: 100%;" data-imgfileid="100011671" src="https://wechat2rss.xlab.app/img-proxy/?k=32460841&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnxiaib1aoTRxhTt7NHvtbJZ3zBCxRKSibFamGn5WN2gxvKAo3Yonso5cKeMvAE6ibHKs7yoZjbCUqueibgKBCJ1Te2wUK6axia5DpUWs%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span><figcaption style="margin-top: 5px;text-align: justify;color: rgb(136, 136, 136);font-size: 14px;"><span leaf="">不同框架与安全模式下的不安全行为率 (UBR) 对比条形图。在 DeepSeek-Chat 和 DeepSeek-Reasoner 模型下，SafeHarness 在所有框架中的不安全行为率均为最低。</span></figcaption></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">实验结果令人瞩目：</span></strong><span leaf="">与无保护的基线相比，SafeHarness 使智能体的 不安全行为率（UBR）平均降低了约 38%，攻击成功率（ASR）平均降低了 42%。特别是在应对极具挑战性的“复合攻击（Composite attacks）”时，层间协同反馈展现出了外部独立防御无法比拟的优势。</span></p><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" alt="多指标雷达图显示，SafeHarness 在显著提升攻击防御成功率的同时，完美保持了任务完成率。" class="rich_pages wxw-img" data-ratio="0.4527777777777778" data-type="png" data-w="1080" style="display: block;margin: 0px auto;max-width: 100%;" data-imgfileid="100011674" src="https://wechat2rss.xlab.app/img-proxy/?k=72a885d6&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnxceOZR2XQpta7eJvce7FENpR3cH7q9zmx0R0mubqKoQUqib6OySsymgnFoeg8EGOCjYvZG9h0DhTvmiaxvKqL14h5ZJvAw4GdFI%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span><figcaption style="margin-top: 5px;text-align: justify;color: rgb(136, 136, 136);font-size: 14px;"><span leaf="">多指标雷达图显示，SafeHarness 在显著提升攻击防御成功率的同时，完美保持了任务完成率。</span></figcaption></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">真正的“六边形战士”：</span></strong><span leaf="">尤为重要的是，如此大幅度的安全提升并没有以牺牲智能体的核心能力为代价。雷达图清晰显示，在安全防线向外大幅扩张的同时，智能体的核心 </span><strong style="font-weight: bold;color: black;"><span leaf="">任务完成率（TCR）</span></strong><span leaf="">几乎零损耗（波动仅 1-2%），系统依然极好地保留了核心任务的实用性（Task Utility） ，打破了“越安全越难用”的魔咒 。</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">6. 案例解析：当智能体遭遇“记忆注入”</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">仅仅看数字不够直观？让我们看看 SafeHarness 是如何在实战中化险为夷的。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">场景：</span></strong><span leaf="">攻击者发起“记忆注入攻击（A4）”，在对话历史中伪造了虚假记录，让智能体误以为用户之前已经授予了它最高权限，可以随意进行破坏性操作（如删除 </span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">/root</span></code><span leaf=""> 目录）。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">❌ </span><strong style="font-weight: bold;color: black;"><span leaf="">无保护的智能体：</span></strong><span leaf="">轻信了伪造的授权记录，直接执行了 </span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">sudo rm -rf /root</span></code><span leaf=""> 命令，导致灾难性的完全不安全行为。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">✅ </span><strong style="font-weight: bold;color: black;"><span leaf="">装备了 SafeHarness 的智能体：</span></strong></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">首先，系统检测到了未经验证的记忆源。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">随后，跨层熵监视器敏锐地察觉到异常，主动将 L2 层的验证级别提升至更严格的 Tier 2/Tier 3。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">L2 层在接连拦截了 4 次危险的工具调用后，通过 Tier 3 级别的因果分析，确诊这是一次注入攻击。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">最终，L4 层触发紧急响应，将会话回滚到上一个安全的检查点。</span></p></li></ol><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" alt="SAFEHARNESS成功拦截并回滚记忆注入攻击的完整判定链条。" class="rich_pages wxw-img" data-ratio="0.6805555555555556" data-type="jpeg" data-w="1080" style="display: block;margin: 0px auto;max-width: 100%;" data-imgfileid="100011676" src="https://wechat2rss.xlab.app/img-proxy/?k=8da25d2c&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnwb3dxQTibl8riavwDYAR01B7YoohkTbEIo9tOreN12EQfOUPFncibn7EkMLMbHINGuy3D5yaictgoJVhwxoHHym9SL20OyciaAibLEE%2F640%3Fwx_fmt%3Djpeg%26from%3Dappmsg"/></span><figcaption style="margin-top: 5px;text-align: justify;color: rgb(136, 136, 136);font-size: 14px;"><span leaf="">SAFEHARNESS成功拦截并回滚记忆注入攻击的完整判定链条。</span></figcaption></figure><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">7. 总结与展望</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">SafeHarness 证明了：将安全机制直接整合到 Harness 架构设计中不仅是可行的，而且是极其有效的。它</span><strong style="font-weight: bold;color: black;"><span leaf="">将安全防线从“外围拦截”深入到了智能体“思考和行动的神经中枢”</span></strong><span leaf="">，为未来更加自主、强大的 AI 智能体在医疗、金融、软件工程等高价值领域的安全落地，奠定了坚实的架构基础。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">欢迎阅读原论文并访问开源代码库</span></strong><span leaf="">，为您的 AI 智能体穿上这件“防弹衣”！</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;font-style: italic;text-decoration: underline;"><a href="https://github.com/liu-yang-maker/SafeHarness" target="_blank">https://github.com/liu-yang-maker/SafeHarness</a></span></span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf=""><span textstyle="" style="font-size: 18px;font-weight: bold;">作者团队介绍</span></span></p><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">论文团队来自</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">中国科学院信息工程研究所第四研究室</span></code><span leaf="">。团队长期从事</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">信息内容安全、大语言模型安全、智能体安全</span></code><span leaf="">等方面的研究，在TPAMI、TKDE、TIFS、ICML、KDD、ACL等CCF-A类期刊与会议上发表学术论文60余篇，相关工作入选最佳论文奖（PAKDD 2020）和最佳论文提名奖（ICDM 2021、WWW 2021）。</span></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://github.com/liu-yang-maker/SafeHarness">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=4a0d2333&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495331%26idx%3D1%26sn%3Dd17385aa84b1614fd8e41d291ec61f2c">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Mon, 11 May 2026 16:21:00 +0800</pubDate>
    </item>
    <item>
      <title>清华大学 | 仅用1.0‰样本训练：基于跨模态特征融合的恶意流量检测</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495303&amp;idx=1&amp;sn=9b07e46bac2b42d3c21192726d1590dc</link>
      <description>本文围绕仅用极少量样本完成训练这一目标展开，提出tFusion，将网络流量同时看作包级、流级和主机级多粒度信息的组合，并试图通过跨模态特征融合提升有限样本条件下的检测能力。</description>
      <content:encoded><![CDATA[<p>原创 <span>龙函城</span> <span>2026-04-28 22:07</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=2da73125&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnzquEFJicZq4CxDYBXWrgKRfOicFT9rPqfnDkic5IKoFE0YIU1Xibb0jyGaAHiaTdUORDCysib7Ficl50RoXUnOoQNfbxsJliaaML2cSicw%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>本文围绕仅用极少量样本完成训练这一目标展开，提出tFusion，将网络流量同时看作包级、流级和主机级多粒度信息的组合，并试图通过跨模态特征融合提升有限样本条件下的检测能力。</p>
  <div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011640" data-ratio="0.2675925925925926" data-type="png" data-w="1080" style="display: block;margin: 0px auto;max-width: 100%;width: 558px;height: auto !important;" src="https://wechat2rss.xlab.app/img-proxy/?k=817c9f24&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnxZfvNhwK4WExvOmK1bHgNF7qJicj8DyvlMHaepaLHQxoPGmn0QbOU2nSaoeYnoSFjvibL0ciceaVyATrQCFicaIvFziacfYHYibORss%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><blockquote style="border: none;display: block;font-size: 0.9em;overflow: auto;overflow-scrolling: touch;border-left: 3px solid rgba(0, 0, 0, 0.4);background: rgba(0, 0, 0, 0.05);color: #6a737d;padding-top: 10px;padding-bottom: 10px;padding-left: 20px;padding-right: 10px;margin-bottom: 20px;margin-top: 20px;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;color: black;line-height: 26px;text-align: justify;"><em style="font-style: italic;color: black;"><span leaf="">原文标题：Training with Only 1.0 ‰ Samples: Malicious Traffic Detection via Cross-Modality Feature Fusion</span></em><br/><em style="font-style: italic;color: black;"><span leaf="">原文作者：Chuanpu Fu, Qi Li, Elisa Bertino, Ke Xu</span></em><br/><em style="font-style: italic;color: black;"><span leaf="">发表会议：CCS&#39;25</span></em><br/><em style="font-style: italic;color: black;"><span leaf="">笔记作者：龙函城</span></em><br/><em style="font-style: italic;color: black;"><span leaf="">主编：黄诚@安全学术圈</span></em></p></blockquote><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span><span leaf="">研究概述</span></span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">传统基于机器学习的恶意流量检测方法虽然能够较好识别未知攻击，但通常依赖大规模标注数据集。一旦系统需要部署到新的网络环境中，研究人员往往必须重新采集并标注大量本地流量样本，这会带来很高的人力与时间成本。本文正是针对这一问题展开研究，其核心目标是在全新网络环境中，仅利用极少量训练样本，仍然保持较高的恶意流量检测精度。论文将这一问题概括为在极小规模数据集条件下训练，并试图突破现有方法对大规模训练数据的依赖。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">为了解决这一问题，作者提出了tFusion。与已有方法通常只从单一粒度建模流量不同，tFusion将网络流量视为一种多模态数据，同时从包级、流级和主机级三个视角提取异构特征，并通过跨模态注意力机制对不同粒度的信息进行关联建模。作者认为，现有基于单一包级、流级或主机级特征的检测方式容易造成信息损失，因此需要大量样本来拟合复杂决策边界；而通过跨粒度融合，可以从有限数据中提取更丰富的判别信息。</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011639" data-type="png" style="display: block;margin: 0px auto;max-width: 100%;width: 515px;height: auto !important;" data-ratio="0.33796296296296297" data-w="1080" src="https://wechat2rss.xlab.app/img-proxy/?k=b075c29a&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnzL9Hkl8HTPNoCpWdSHlKNfOrkz870MfwcDribAtPibl7R8Wib3rVDPpl3yOHC540vsTZ0JB3O0nibNwef0X2C9FBaQsjblUnOMflY%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">图 1 tFusion恶意流量检测系统架构</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如图1所示，tFusion 的整体流程主要包括四个阶段。首先，系统通过时间感知的数据包序列嵌入对数据包序列进行时间感知嵌入，在编码包长信息和到达时间间隔信息的基础上，利用自注意力机制提取细粒度的包级时序特征。其次，系统进入跨模态注意力特征融合阶段，提取流级统计特征和主机级交互特征，并以主机级特征作为query、流级特征作为key、包级特征作为value，构建跨模态注意力关系模型，实现多粒度特征融合。随后，论文进一步设计了拓扑驱动的对比学习，利用大规模无标签互联网流量对模型进行预训练，使其在不同网络拓扑环境下获得更强的泛化表示能力。最后，在部署阶段，系统只需从目标网络中随机抽取极少量流量样本，在完成人工标注后，即可借助轻量级机器学习检测模块训练轻量级的监督或无监督模型，从而完成实际检测任务。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从实验结果分析可以看出，tFusion在有限样本场景下具有较强的适应能力。无论是在11个公开数据集上的对比实验，还是在真实机构网络中的部署测试中，该方法都表现出了较好的检测效果。特别是在真实环境下，仅对1.0‰的流量样本进行人工标注后，系统仍然达到了99.82%的检测准确率。这表明，tFusion通过联合建模不同粒度的流量特征，能够在样本规模很小的情况下仍保持较高的检测性能，对于实际网络中的低成本部署具有一定参考价值。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span><span leaf="">贡献分析</span></span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点1</span></strong><span leaf=""> ：提出基于跨模态注意力的多粒度流量特征关联方法</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文将网络流量视为由包级、流级和主机级等不同粒度特征共同构成的多模态数据，并利用跨模态注意力机制对这些异构特征进行关联建模。该方法能够在训练样本十分有限的情况下，更充分地挖掘流量中的判别信息，从而提升恶意流量检测效果。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点2</span></strong><span leaf=""> ：设计时间感知的数据包序列嵌入方法</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">为了更好地提取细粒度包级时序特征，论文提出了时间感知的数据包序列嵌入方法。该方法不仅考虑数据包在序列中的位置关系，还同时编码包长和到达时间间隔信息，从而增强了对恶意流量中时序模式和行为结构的刻画能力。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点3</span></strong><span leaf=""> ：设计跨注意力机制实现细粒度与粗粒度特征融合</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">在包级特征提取的基础上，论文进一步设计了cross-attention机制，用于融合细粒度的数据包特征与粗粒度的流级、主机级统计特征。模型能够在不同粒度之间建立更有效的相关性表示，而不只是简单地进行特征拼接，从而提升了特征融合质量。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点4</span></strong><span leaf=""> ：提出基于拓扑驱动对比学习的无标签预训练策略</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">为减少对标注数据的依赖，论文利用大规模无标签互联网流量对注意力模型进行预训练，并提出拓扑驱动的对比学习方法。该策略能够在不同网络拓扑环境下学习更具泛化性的流量表示，为后续在新网络中仅使用极少量样本进行训练奠定基础。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点5</span></strong><span leaf=""> ：实现tFusion原型系统并完成多环境验证</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文最终实现了tFusion原型系统，并在多种网络环境中对其准确性和效率进行了验证。实验结果表明，该方法不仅在多个公开数据集上表现优于已有方法，而且在真实机构网络中也能够以极少量标注样本实现较高精度检测，说明其具有较好的实际应用潜力。</span></p></div></li></ul><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span><span leaf="">代码分析</span></span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">代码链接</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf=""><a href="https://github.com/fuchuanpu/TFusion" target="_blank">https://github.com/fuchuanpu/TFusion</a></span></p></li></ul><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">使用类库分析</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">该项目整体采用 Python 实现，辅以少量脚本文件，整体属于较为典型的论文实验原型代码。从仓库组织方式来看，代码主要围绕数据处理、模型训练、特征融合、检测器构建以及实验复现等环节展开。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码实现难度与工作量评估</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从实现角度来看，该项目难度属于中等偏上。它的实现重点不在于某一个单独模型特别复杂，而在于需要把多模态特征构造、对比学习预训练、跨模态特征融合以及下游检测器训练等多个环节有效串联起来。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码关键实现功能</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（1）多粒度流量特征构造模块。该模块主要负责对输入流量数据进行组织与处理，提取不同粒度的特征表示，包括细粒度的序列行为特征以及较粗粒度的统计与交互特征，为后续的多模态建模提供统一输入。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（2）时间感知序列表示学习模块。该模块是对数据包序列中的长度、顺序以及到达时间间隔等信息进行联合编码，从而提取更能反映时序模式的包级特征表示。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（3）跨模态特征融合模块。该模块主要用于将包级、流级和主机级等不同粒度特征进行关联建模和融合表示，目标是通过建模不同粒度特征之间的关系，获得更具判别力的流量表示。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（4）对比学习预训练模块。该模块主要承担预训练任务，通过大规模无标签流量数据学习更具泛化能力的特征表示，是整个方法能够在小样本场景下保持较好性能的重要支撑部分。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（5）轻量级检测器训练与测试模块。在获得融合后的流量表示后，系统会进一步训练轻量级监督或无监督检测模型，并完成测试评估。</span></p></div></li></ol><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span><span leaf="">论文点评</span></span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">总体来看，这篇论文的价值在于它抓住了一个非常现实的问题：恶意流量检测方法在实验环境中往往依赖大规模标注数据集，但在真正部署到新网络时，重新采集和标注本地流量的成本往往很高。本文围绕仅用极少量样本完成训练这一目标展开，提出tFusion，将网络流量同时看作包级、流级和主机级多粒度信息的组合，并试图通过跨模态特征融合提升有限样本条件下的检测能力。这一问题设定具有较强的实际意义，也使论文区别于许多主要关注常规大样本训练场景的恶意流量检测研究。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，这篇论文也有一定局限。首先，tFusion 的优势主要体现在有限样本条件下的检测能力提升，因此其贡献重点更多在于训练范式和特征融合思路，而不是提出一种全新的检测任务；其次，论文虽然在真实网络中做了部署实验，但整体验证仍然主要依赖公开数据集和人工构造攻击场景，距离更复杂、长期开放环境下的持续验证还有一定距离。总体而言，这篇论文的主要意义在于，它为恶意流量检测中的小样本训练问题提供了一种较清晰、可实现的解决思路，对于后续研究有限标注条件下的网络安全检测方法具有较强参考价值。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span><span leaf="">论文文献</span></span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">Fu, et al. &#34;Training with Only 1.0‰ Samples: Malicious Traffic Detection via Cross-Modality Feature Fusion.&#34; Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security. 2025.</span></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link mp_article_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp"><mp-common-profile class="custom_select_card mp_profile_iframe mp_common_widget" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA=="></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://dl.acm.org/doi/pdf/10.1145/3719027.3765143">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=9aaf0fca&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495303%26idx%3D1%26sn%3D9b07e46bac2b42d3c21192726d1590dc">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Tue, 28 Apr 2026 22:07:00 +0800</pubDate>
    </item>
    <item>
      <title>IEEE成都分会 | AI智能体在学术研究中的应用</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495302&amp;idx=1&amp;sn=0e6ee52b5f1994f8ab0a1e0bba37e3ab</link>
      <description>IEEE 成都分会 | 博士生与青年专业人员论坛</description>
      <content:encoded><![CDATA[<p>原创 <span>小编</span> <span>2026-04-25 09:00</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=4f282145&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnwWR4zuXpelssp4ymSjRMzDqNEgibbfKg04PNtrAn0VwiaJxsFcaRIaZCaPGvuDXEg18ha739X479maCPIeQLCoBu8MvAVUuMHzs%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>IEEE 成都分会 | 博士生与青年专业人员论坛</p>
  <div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011653" data-ratio="0.27870370370370373" style="display: block;margin: 0 auto;max-width: 100%;" data-type="png" data-w="1080" src="https://wechat2rss.xlab.app/img-proxy/?k=1747049d&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnyAKI06iaQiaYKHCjHeu9fhppbibticIBxBehHF81QxgiaQIDtN1kDq2QEH1blaMkDd3eZ1jvUibqAyFtaV6icqXxzWkDEzyfZalTDNYE%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">讲座标题</span></strong><span leaf="">：AI 智能体简介及其在学术研究中的应用</span><span leaf=""><br/></span><strong style="font-weight: bold;color: black;"><span leaf="">讲座时间</span></strong><span leaf="">：</span><code style="font-size: 14px;word-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0 2px;color: #1e6bb8;background-color: rgba(27,31,35,.05);font-family: Operator Mono, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">2026年4月28日（周二）上午10:00-11:00 </span></code><span leaf=""><br/></span><strong style="font-weight: bold;color: black;"><span leaf="">线下地点</span></strong><span leaf="">：中国成都，四川大学望江校区</span><span leaf=""><br/></span><strong style="font-weight: bold;color: black;"><span leaf="">腾讯会议</span></strong><span leaf="">：</span><code style="font-size: 14px;word-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0 2px;color: #1e6bb8;background-color: rgba(27,31,35,.05);font-family: Operator Mono, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">803-986-541</span></code><span leaf=""><br/></span><strong style="font-weight: bold;color: black;"><span leaf="">演讲人</span></strong><span leaf="">：王少通</span><span leaf=""><br/></span><strong style="font-weight: bold;color: black;"><span leaf="">主持人</span></strong><span leaf="">：周颖杰（四川大学教授）</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">1、摘要</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">大型语言模型催生了越来越多的基于 AI 智能体的工具。一些用户主要通过聊天界面与大型语言模型交互，而另一些用户可能仍在探索如何将 AI 智能体深度集成到他们的研究工作流程中。本次演讲将探讨 AI 智能体如何帮助提高学术研究的效率。它将从 AI 智能体的定义开始，并介绍它们如何协助一系列任务，如</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">文献综述、编码、编辑、演示和文档格式化</span></code><span leaf="">。本次演讲将对当前主流 AI 智能体的格局进行分类，并介绍一些有用的技术，如提示工程、上下文工程、MCP、技能和多智能体协作。最后，将讨论基于智能体的最新研究系统、研究工作流程中的新兴趋势以及相关的见解。</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">2、演讲人简介</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">王少通目前正在四川大学计算机科学学院攻读博士学位。他毕业于成都理工大学，获得信息与计算科学学士学位。目前的研究兴趣包括 AI 智能体和网络安全。</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">3、直播预约</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">✅ 腾讯会议号：803-986-541</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">✅ 请扫描下方二维码预约 4 月 28 日（周二）上午的活动直播（微信视频号）：</span></p><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.7936893203883495" data-type="png" data-w="824" style="display:block;margin:0px auto;max-width:100%;width:186px;height:148px;" data-imgfileid="100011648" src="https://wechat2rss.xlab.app/img-proxy/?k=0e1b9db4&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnzicGPnZAjM7dXRur5Y9QkobOArvHZNwGhfFr0lv5YpqpjUJJrWofKE7vJIsV8uQP2siaJHRfiaibdYpFlpKjTYfIx7kG3qicN5ZMl4%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>



<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=cedbdc61&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495302%26idx%3D1%26sn%3D0e6ee52b5f1994f8ab0a1e0bba37e3ab">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Sat, 25 Apr 2026 09:00:00 +0800</pubDate>
    </item>
    <item>
      <title>鹏城实验室-西电计算机｜课题组联合培养博士生招生</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495292&amp;idx=1&amp;sn=e5e05e13f04b1207065dd791afbadfcf</link>
      <description>鹏城-西电计算机联培｜沈玉龙&amp;谢国锐老师招收AI+计算机网络方向博士生</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-24 08:40</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=173e5abc&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnyxlP8uliaBpbbVpQE6IAKSTTRDnohaop30ySXSSGbY4jeDL8EkKQz5JxpqehMpJVfVYTGPgjISXk71pAPib5G3y25N5302bYficw%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>鹏城-西电计算机联培｜沈玉龙&谢国锐老师招收AI+计算机网络方向博士生</p>
  <div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><h1 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 24px;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011636" data-ratio="0.42024539877300615" style="display: block;margin: 0 auto;max-width: 100%;" data-type="png" data-w="978" src="https://wechat2rss.xlab.app/img-proxy/?k=3aca912e&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnxnrYQNmX7IXgicibB22mumoPAW4h8n2xibhfFfZFliciaar1YwWNaD41l2UlSOicvTJdbicUK6ib0zHz5MpicHbLHxIQZJGeqOEwK1sXQk%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></h1><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">一、导师介绍</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">沈玉龙</span></strong><span leaf="">（<a href="https://faculty.xidian.edu.cn/SYL4/zh_CN/index.htm），西安电子科技大学“华山学者”领军教授，博士/硕士研究生导师，国家级人才，陕西省网络与系统安全重点实验室主任，教育部创新团队的骨干成员，综合业务网国家重点实验室信息安全中心固定研究人员。长期从事下一代网络安全、云计算与数据安全等研究。近年来主持和承担了国家重点研发计划、国家自然科学基金重点项目、863、973等20余项，以主要完成人获国家技术发明二等奖2项，省部级一等奖7项。研发的“秦盾”云加密数据库系统获中国科协创新创业大赛金奖（16/1800+）和最佳商业投资价值奖（2/1800+），获2018" target="_blank">https://faculty.xidian.edu.cn/SYL4/zh_CN/index.htm），西安电子科技大学“华山学者”领军教授，博士/硕士研究生导师，国家级人才，陕西省网络与系统安全重点实验室主任，教育部创新团队的骨干成员，综合业务网国家重点实验室信息安全中心固定研究人员。长期从事下一代网络安全、云计算与数据安全等研究。近年来主持和承担了国家重点研发计划、国家自然科学基金重点项目、863、973等20余项，以主要完成人获国家技术发明二等奖2项，省部级一等奖7项。研发的“秦盾”云加密数据库系统获中国科协创新创业大赛金奖（16/1800+）和最佳商业投资价值奖（2/1800+），获2018</a> CCTV中国十大创业榜样（高校首次）。发表SCI/EI论文100余篇，Google scholar他引2000余次，出版专著7部；授权发明专利70余项，PCT专利5项，已转让专利13项；登记软著40多项；制定中国通信/电子行业标准5项。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">谢国锐</span></strong><span leaf="">（<a href="https://xgr19.github.io/），2024年于清华大学获博士学位，本科毕业于中山大学（2019年）。现任鹏城国家实验室助理研究员，鹏城-清华/中大/西电等高校联培博士生导师（清华智能网络团队，https://smartinternet.group/），CCF分布式计算与系统、互联网专委会执委，并担任CoNEXT、JSAC、WWW等国际会议与期刊的程序委员或审稿人。研究方向为可编程网络、AI/LLM的网络应用及智能流量分析系统，在NSDI、INFOCOM、ToN、TDSC、TC等CCF-A类会议与期刊上发表论文20余篇。曾获国家留学基金委资助、腾讯青云计划。作为核心成员参与院士工作室多项千万级重大项目，成果获国内与国际PCT专利8项。协助指导的硕士生曾入选大厂人才计划、优秀毕业生。" target="_blank">https://xgr19.github.io/），2024年于清华大学获博士学位，本科毕业于中山大学（2019年）。现任鹏城国家实验室助理研究员，鹏城-清华/中大/西电等高校联培博士生导师（清华智能网络团队，https://smartinternet.group/），CCF分布式计算与系统、互联网专委会执委，并担任CoNEXT、JSAC、WWW等国际会议与期刊的程序委员或审稿人。研究方向为可编程网络、AI/LLM的网络应用及智能流量分析系统，在NSDI、INFOCOM、ToN、TDSC、TC等CCF-A类会议与期刊上发表论文20余篇。曾获国家留学基金委资助、腾讯青云计划。作为核心成员参与院士工作室多项千万级重大项目，成果获国内与国际PCT专利8项。协助指导的硕士生曾入选大厂人才计划、优秀毕业生。</a></span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">二、招生方向及政策</span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">招生方向为计算机网络（AI4Network），特别欢迎有AI/LLM基础、可编程网络、网络安全等研究领域背景的学生联系</span></p></li></ul><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">申请-考核制博士生【1名】，官网报名系统截止时间为</span><strong style="font-weight: bold;color: black;"><span leaf="">4月28日</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">鹏城实验室学生培养津贴丰厚，原则上学生完成课程学习之后在鹏城实验室项目团队及导师组的指导下参与鹏城实验室重大攻关任务，学位论文选题须来源于实验室重大攻关任务</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">详细要求可查看鹏城官网（<a href="https://yzw.pcl.ac.cn/home/homeDetail?id=1946152383544987650）及腾讯文档（https://docs.qq.com/doc/DRkFDb0dHbXdKY0R6）" target="_blank">https://yzw.pcl.ac.cn/home/homeDetail?id=1946152383544987650）及腾讯文档（https://docs.qq.com/doc/DRkFDb0dHbXdKY0R6）</a></span></p></li></ul><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">三、招生要求及联系方式</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">非常欢迎对AI+网络研究有兴趣的学生联系，满足以下条件的加分：</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">✅ 自驱力强、沟通能力强</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">✅ 有一定Coding和系统实现能力</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">感兴趣的同学以“鹏城-西电联培博申请”为邮件同时联系沈老师课题组（cyjiao@xidian.edu.cn）和谢老师（xiegr@pcl.ac.cn），并附上个人简历。</span></strong></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">四、研究所/学校介绍</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">鹏城国家实验室</span></strong><span leaf="">位于深圳市，是中央批准成立的新型科研机构，作为国家战略科技力量的重要组成部分，实验室聚焦宽带通信、网络智能等国家重大战略任务以及粤港澳大湾区、中国特色社会主义先行示范区建设的长远目标，按照“四个面向”的要求，开展领域内重大科学问题研究。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">西安电子科技大学计算机科学与技术学院</span></strong><span leaf="">（国家示范性软件学院）始于1958年中央军委批复设立的导弹系统专用计算机专业。现设有4个系、8个研究所、1个实践教学中心，建有“计算机网络与信息安全”国家级实验教学示范中心、国家高等学校创新引智计划“大数据基础理论与智能分析学科创新引智基地”、“区块链技术应用与评测”教育部工程研究中心等10个省部级以上科研平台。</span></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>



<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=be17cfc3&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495292%26idx%3D1%26sn%3De5e05e13f04b1207065dd791afbadfcf">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Fri, 24 Apr 2026 08:40:00 +0800</pubDate>
    </item>
    <item>
      <title>行业会议 | 第七届网络空间安全前沿论坛（常安论坛详细全日程）</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495279&amp;idx=1&amp;sn=5620974b374e7d9da2d34f5af5f9dbb3</link>
      <description>常安论坛详细全日程发布</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-23 11:45</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=4cbc060b&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnxa5HzPvicKyrxdYCC5Xs2qW2RvyaV7foTVKl0EV1zgobLguAakkXUtEZFic4KZwgytcYCqVdVIkabicRNWYFW2xIboEnVnIJGiaas%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>常安论坛详细全日程发布</p>
  <p style="text-align: center;" nodeleaf=""><img class="rich_pages wxw-img js_insertlocalimg" data-aistatus="1" data-imgfileid="100011628" data-ratio="16.186" data-s="300,640" type="block" data-type="webp" data-w="1000" src="https://wechat2rss.xlab.app/img-proxy/?k=ba70ac53&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnyeuWL5EKjQh7EV3pHdllXgJP7lBdbMiaibgqafMaL5LoHaHPhMUUN5asFZetWOL7YzGxauNQ0icDiaZtEDICAAGGHBwUmZorERgWQ%2F640%3Fwx_fmt%3Dwebp%26from%3Dappmsg"/></p><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="letter-spacing: 0.578px;border-style: solid;border-right-width: 0px;border-bottom-width: 0px;border-left-width: 0px;border-color: rgba(0, 0, 0, 0.1);transform-origin: 0px 0px;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>



<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=9e28eb71&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495279%26idx%3D1%26sn%3D5620974b374e7d9da2d34f5af5f9dbb3">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Thu, 23 Apr 2026 11:45:00 +0800</pubDate>
    </item>
    <item>
      <title>会议征稿 | EXPRESS 2026</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495274&amp;idx=1&amp;sn=e544afc1fff1200b156bc89a1367021d</link>
      <description>截稿日期：2026年6月14日</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-22 11:31</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=5b4bd7cf&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnwpiaRGib6nycQep9QgG6U2JQwS9TdBzg21Iu3jOW5hU84lLEsbMFv91FG5qlqjeyhxtsjZ72f0icSBVSibSiaCSJYuQ8psx4KpfmLs%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>截稿日期：2026年6月14日</p>
  <p style="text-align: center;margin-bottom: 0px;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img js_insertlocalimg" data-ratio="0.4255555555555556" data-s="300,640" data-type="jpeg" data-w="900" style="width: 578px;height: 246px;" data-cropselx1="0" data-cropselx2="578" data-cropsely1="0" data-cropsely2="244" src="https://wechat2rss.xlab.app/img-proxy/?k=e0f42fc8&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F6Dibw6L070WFQBicGiaWuqeNibz9LVpfzMvHfO3CW2NwCGEHtibXTZf6Pm5ndjZJibw226glYGjhC8wPPupX0cIaiaVNg%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">当代软件系统正经历一场深刻的代际跃迁。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如果说过去十年，我们见证了软件架构向</span><strong style="font-weight: bold;color: black;"><span leaf="">云原生</span></strong><span leaf="">的全面转型；那么随着生成式人工智能技术的爆发，我们正加速迈入</span><strong style="font-weight: bold;color: black;"><span leaf="">AI原生</span></strong><span leaf="">的新时代。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">在这一变革中，AI不再是外挂的辅助工具，而是正在演变为软件架构的核心驱动力、代码的生产者以及系统决策的“大脑”。然而，这种转变在赋予系统前所未有的自主性与智能化的同时，也带来了全新的复杂性与不确定性。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">核心挑战：</span></strong><span leaf=""> 如何在从云原生向AI原生演进的过程中，系统性保障软件系统的</span><strong style="font-weight: bold;color: black;"><span leaf="">正确性、可解释性与可信赖性</span></strong><span leaf="">？这已成为软件工程与DevOps领域面临的最紧迫挑战。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">继往开来：</span></strong><span leaf=""> 继首届 EXPRESS 2025 于挪威特隆赫姆成功探索“复杂软件系统与大语言模型融合的可解释性”之后，</span><strong style="font-weight: bold;color: black;"><span leaf="">第二届 EXPRESS 研讨会</span></strong><span leaf=""> 将进一步深化议题，立足 </span><strong style="font-weight: bold;color: black;"><span leaf="">AI驱动的软件研运一体化（AI辅助DevOps）</span></strong><span leaf="">，探讨在这一代际切换中，如何构建既具备强大智能能力、又拥有坚实可信底座的软件系统。</span></p><hr style="height: 1px;margin: 10px 0px;border-width: 1px medium medium;border-style: solid none none;border-color: black currentcolor currentcolor;border-image: initial;"/><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">会议信息</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">EXPRESS 2026</span></strong><span leaf=""> 将与美国软件工程领域两大顶级会议联合举办： </span><strong style="font-weight: bold;color: black;"><span leaf="">SPLASH 2026 &amp; ISSTA 2026</span></strong><span leaf=""> ，</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">时间：</span></strong><span leaf=""> 2026年 10月 3日 - 2026年 10 月 9日</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">地点：</span></strong><span leaf=""> 美国加利福尼亚州奥克兰市</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">联合会议背景介绍：</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">ISSTA</span></strong><span leaf=""> (ACM SIGSOFT International Symposium on Software Testing and Analysis)：软件测试与分析领域公认的权威性顶级会议之一。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">SPLASH</span></strong><span leaf=""> (ACM SIGPLAN conference on Systems, Programming, Languages, and Applications)：编程语言与软件工程领域的顶级国际会议之一。</span></p></li></ul><hr style="height: 1px;margin: 10px 0px;border-width: 1px medium medium;border-style: solid none none;border-color: black currentcolor currentcolor;border-image: initial;"/><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">征稿主题</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">我们诚邀包括但不限于以下方向的原创工作投稿，特别欢迎从严谨理论基础、工程方法、工具实现到真实案例与经验报告等不同角度的交叉研究与多学科合作：</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">软件及智能体系统的可靠性、安全性与可信性</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">面向可信软件的分析、测试与（形式）验证技术</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">面向软件系统的可信 AI，以及支撑可信 AI 的软件系统基础设施与工具链</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">机器学习模型与软件系统的可解释性方法与实践</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">基于大语言模型的验证、测试与分析技术及其可解释性与可靠性评估</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">软件系统依赖关系与复杂性的发现、分析与挖掘</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">面向增强可解释性的软硬件系统可视化技术与工具</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">运行时可观测性、运行时分析、监控及错误恢复机制</span></strong></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">我们特别鼓励以下类型的工作：</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">兼具严谨理论基础与工程可落地性的方法与框架</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">面向真实复杂场景的工具实现、平台方案与实践经验总结</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">融合软件工程、机器学习、运筹优化、系统架构等多学科视角的交叉研究</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">从“云原生”向“AI 原生”演进过程中的架构演化、流程改造与组织协同实践</span></strong></p></li></ul><hr style="height: 1px;margin: 10px 0px;border-width: 1px medium medium;border-style: solid none none;border-color: black currentcolor currentcolor;border-image: initial;"/><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">投稿指引</span></h3><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">会议主页：</span></strong><span leaf=""><a href="https://conf.researchr.org/home/splash-issta-2026/express-2026" target="_blank">https://conf.researchr.org/home/splash-issta-2026/express-2026</a></span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">投稿链接：</span></strong><span leaf=""><a href="https://express26.hotcrp.com" target="_blank">https://express26.hotcrp.com</a></span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">截稿日期：</span></strong><strong style="font-weight: bold;color: black;"><span leaf="">2026年6月14日</span></strong></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">投稿形式：</span></strong><span leaf=""> 本届研讨会接收不同类型的论文，包括：</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Full Papers (完整技术论文)：</span></strong><span leaf=""> 最多 8 页</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Work-in-Progress Papers (研究进展论文)：</span></strong><span leaf="">最多 4 页</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Industry &amp; Tool Papers (工业界与工具论文)：</span></strong><span leaf=""> 最多 4 页</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><em style="font-style: italic;color: black;"><span leaf="">(所有论文必须为英文，使用 ACM Primary Article Template 格式，页数限制不包含参考文献)</span></em></p><hr style="height: 1px;margin: 10px 0px;border-width: 1px medium medium;border-style: solid none none;border-color: black currentcolor currentcolor;border-image: initial;"/><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">组织机构</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">组织委员会</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">刘圃卓</span></strong><span leaf=""> | 蚂蚁集团</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">贺品嘉</span></strong><span leaf=""> | 香港中文大学（深圳）</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">张洪宇</span></strong><span leaf=""> | 重庆大学</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">南雨宏</span></strong><span leaf=""> | 中山大学</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">指导委员会</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">狄 鹏</span></strong><span leaf=""> | 蚂蚁集团</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">姜 宇</span></strong><span leaf=""> | 清华大学</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">薛京灵</span></strong><span leaf=""> | 新南威尔士大学 (UNSW)</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">周智泉</span></strong><span leaf=""> | 蔚来汽车</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">程序委员会</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Yi Chen</span></strong><span leaf="">｜University of Hong Kong</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Wenrui Diao</span></strong><span leaf="">｜Shandong University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Cheng Huang</span></strong><span leaf="">｜Sichuan University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Yan Jia</span></strong><span leaf="">｜Nankai University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Siyuan Li</span></strong><span leaf="">｜Shandong University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Hongyu Liu</span></strong><span leaf="">｜Huawei</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Yepang Liu</span></strong><span leaf="">｜Southern University of Science and Technology</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Yingwei Ma</span></strong><span leaf="">｜Alibaba Group</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Sagar Mohan</span></strong><span leaf="">｜Northeastern University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Chuan Qin</span></strong><span leaf="">｜Nanyang Technological University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Xiaoxue Ren</span></strong><span leaf="">｜Zhejiang University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Qingkai Shi</span></strong><span leaf="">｜Nanjing University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Guannan Wei</span></strong><span leaf="">｜Tufts University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Haoran Yang</span></strong><span leaf="">｜Sichuan University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Zhou Yang</span></strong><span leaf="">｜University of Alberta</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Mingxi Ye</span></strong><span leaf="">｜Sun Yat-sen University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Juan Zhai</span></strong><span leaf="">｜University of Massachusetts at Amherst</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Xiaohan Zhang</span></strong><span leaf="">｜Fudan University</span></p></li></ul><hr style="height: 1px;margin: 10px 0px;border-width: 1px medium medium;border-style: solid none none;border-color: black currentcolor currentcolor;border-image: initial;"/><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">EXPRESS 2026 期待您的投稿，相聚旧金山湾区，共话 AI 原生时代的软件研运一体化的未来！</span></strong></p><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p style="margin-bottom: 0px;color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;font-size: 16px;letter-spacing: 0.544px;text-align: center;caret-color: rgb(0, 0, 0);line-height: 1.75em;"><span style="color: rgb(0, 82, 255);"><strong><span leaf=""><br/></span></strong></span></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://conf.researchr.org/home/splash-issta-2026/express-2026">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=dcc37f17&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495274%26idx%3D1%26sn%3De544afc1fff1200b156bc89a1367021d">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Wed, 22 Apr 2026 11:31:00 +0800</pubDate>
    </item>
    <item>
      <title>【征稿】EXPRESS 2026：从云原生到AI原生，重塑软件系统的可解释性与可靠性</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495270&amp;idx=1&amp;sn=3e7673892424a6d556cd051eeb27e368</link>
      <description>截稿日期： 2026年6月14日</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-18 08:03</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=a6e88efd&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnxHTH4jLnIAmzDdxY7aib9auh0UeQcANvR0ILDLXMOsPibOuDuVteV1ZicuBZx9gqB5uUajlNTtztYyzsGE0ImyWAOFHYx0ibediblQ%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>截稿日期： 2026年6月14日</p>
  <p style="text-align: center;margin-bottom: 0px;"><span leaf=""><img class="rich_pages wxw-img js_insertlocalimg" data-ratio="0.4255555555555556" data-s="300,640" data-type="png" data-w="900" style="width: 578px;height: auto !important;" data-cropselx1="0" data-cropselx2="578" data-cropsely1="0" data-cropsely2="244" data-aistatus="1" src="https://wechat2rss.xlab.app/img-proxy/?k=e0f42fc8&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F6Dibw6L070WFQBicGiaWuqeNibz9LVpfzMvHfO3CW2NwCGEHtibXTZf6Pm5ndjZJibw226glYGjhC8wPPupX0cIaiaVNg%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">背景</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">当代软件系统正经历一场深刻的代际跃迁。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如果说过去十年，我们见证了软件架构向</span><strong style="font-weight: bold;color: black;"><span leaf="">云原生</span></strong><span leaf="">的全面转型；那么随着生成式人工智能技术的爆发，我们正加速迈入</span><strong style="font-weight: bold;color: black;"><span leaf="">AI原生</span></strong><span leaf="">的新时代。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">在这一变革中，AI不再是外挂的辅助工具，而是正在演变为软件架构的核心驱动力、代码的生产者以及系统决策的“大脑”。然而，这种转变在赋予系统前所未有的自主性与智能化的同时，也带来了全新的复杂性与不确定性。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">核心挑战：</span></strong><span leaf=""> 如何在从云原生向AI原生演进的过程中，系统性保障软件系统的</span><strong style="font-weight: bold;color: black;"><span leaf="">正确性、可解释性与可信赖性</span></strong><span leaf="">？这已成为软件工程与DevOps领域面临的最紧迫挑战。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">继往开来：</span></strong><span leaf=""> 继首届 EXPRESS 2025 于挪威特隆赫姆成功探索“复杂软件系统与大语言模型融合的可解释性”之后，</span><strong style="font-weight: bold;color: black;"><span leaf="">第二届 EXPRESS 研讨会</span></strong><span leaf=""> 将进一步深化议题，立足 </span><strong style="font-weight: bold;color: black;"><span leaf="">AI驱动的软件研运一体化（AI辅助DevOps）</span></strong><span leaf="">，探讨在这一代际切换中，如何构建既具备强大智能能力、又拥有坚实可信底座的软件系统。</span></p><hr style="height: 1px;margin: 0;margin-top: 10px;margin-bottom: 10px;border: none;border-top: 1px solid black;"/><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">会议信息</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">EXPRESS 2026</span></strong><span leaf=""> 将与美国软件工程领域两大顶级会议联合举办： </span><strong style="font-weight: bold;color: black;"><span leaf="">SPLASH 2026 &amp; ISSTA 2026</span></strong><span leaf=""> ，</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">时间：</span></strong><span leaf=""> 2026年 10月 3日 - 2026年 10 月 9日</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">地点：</span></strong><span leaf=""> 美国加利福尼亚州奥克兰市</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">联合会议背景介绍：</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">ISSTA</span></strong><span leaf=""> (ACM SIGSOFT International Symposium on Software Testing and Analysis)：软件测试与分析领域公认的权威性顶级会议之一。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">SPLASH</span></strong><span leaf=""> (ACM SIGPLAN conference on Systems, Programming, Languages, and Applications)：编程语言与软件工程领域的顶级国际会议之一。</span></p></li></ul><hr style="height: 1px;margin: 0;margin-top: 10px;margin-bottom: 10px;border: none;border-top: 1px solid black;"/><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">征稿主题</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">我们诚邀包括但不限于以下方向的原创工作投稿，特别欢迎从严谨理论基础、工程方法、工具实现到真实案例与经验报告等不同角度的交叉研究与多学科合作：</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">软件及智能体系统的可靠性、安全性与可信性</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">面向可信软件的分析、测试与（形式）验证技术</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">面向软件系统的可信 AI，以及支撑可信 AI 的软件系统基础设施与工具链</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">机器学习模型与软件系统的可解释性方法与实践</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">基于大语言模型的验证、测试与分析技术及其可解释性与可靠性评估</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">软件系统依赖关系与复杂性的发现、分析与挖掘</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">面向增强可解释性的软硬件系统可视化技术与工具</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">运行时可观测性、运行时分析、监控及错误恢复机制</span></strong></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">我们特别鼓励以下类型的工作：</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">兼具严谨理论基础与工程可落地性的方法与框架</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">面向真实复杂场景的工具实现、平台方案与实践经验总结</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">融合软件工程、机器学习、运筹优化、系统架构等多学科视角的交叉研究</span></strong></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">从“云原生”向“AI 原生”演进过程中的架构演化、流程改造与组织协同实践</span></strong></p></li></ul><hr style="height: 1px;margin: 0;margin-top: 10px;margin-bottom: 10px;border: none;border-top: 1px solid black;"/><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">投稿指引</span></h3><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">会议主页：</span></strong><span leaf=""><a href="https://conf.researchr.org/home/splash-issta-2026/express-2026" target="_blank">https://conf.researchr.org/home/splash-issta-2026/express-2026</a></span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">投稿链接：</span></strong><span leaf=""><a href="https://express26.hotcrp.com" target="_blank">https://express26.hotcrp.com</a></span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">截稿日期：</span></strong><strong style="font-weight: bold;color: black;"><span leaf="">2026年6月14日</span></strong></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">投稿形式：</span></strong><span leaf=""> 本届研讨会接收不同类型的论文，包括：</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Full Papers (完整技术论文)：</span></strong><span leaf=""> 最多 8 页</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Work-in-Progress Papers (研究进展论文)：</span></strong><span leaf="">最多 4 页</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Industry &amp; Tool Papers (工业界与工具论文)：</span></strong><span leaf=""> 最多 4 页</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><em style="font-style: italic;color: black;"><span leaf="">(所有论文必须为英文，使用 ACM Primary Article Template 格式，页数限制不包含参考文献)</span></em></p><hr style="height: 1px;margin: 0;margin-top: 10px;margin-bottom: 10px;border: none;border-top: 1px solid black;"/><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">组织机构</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">组织委员会</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">刘圃卓</span></strong><span leaf=""> | 蚂蚁集团</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">贺品嘉</span></strong><span leaf=""> | 香港中文大学（深圳）</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">张洪宇</span></strong><span leaf=""> | 重庆大学</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">南雨宏</span></strong><span leaf=""> | 中山大学</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">指导委员会</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">狄 鹏</span></strong><span leaf=""> | 蚂蚁集团</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">姜 宇</span></strong><span leaf=""> | 清华大学</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">薛京灵</span></strong><span leaf=""> | 新南威尔士大学 (UNSW)</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">周智泉</span></strong><span leaf=""> | 蔚来汽车</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">程序委员会</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Yi Chen</span></strong><span leaf="">｜University of Hong Kong</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Wenrui Diao</span></strong><span leaf="">｜Shandong University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Cheng Huang</span></strong><span leaf="">｜Sichuan University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Yan Jia</span></strong><span leaf="">｜Nankai University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Siyuan Li</span></strong><span leaf="">｜Shandong University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Hongyu Liu</span></strong><span leaf="">｜Huawei</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Yepang Liu</span></strong><span leaf="">｜Southern University of Science and Technology</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Yingwei Ma</span></strong><span leaf="">｜Alibaba Group</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Sagar Mohan</span></strong><span leaf="">｜Northeastern University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Chuan Qin</span></strong><span leaf="">｜Nanyang Technological University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Xiaoxue Ren</span></strong><span leaf="">｜Zhejiang University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Qingkai Shi</span></strong><span leaf="">｜Nanjing University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Guannan Wei</span></strong><span leaf="">｜Tufts University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Haoran Yang</span></strong><span leaf="">｜Sichuan University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Zhou Yang</span></strong><span leaf="">｜University of Alberta</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Mingxi Ye</span></strong><span leaf="">｜Sun Yat-sen University</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Juan Zhai</span></strong><span leaf="">｜University of Massachusetts at Amherst</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">Xiaohan Zhang</span></strong><span leaf="">｜Fudan University</span></p></li></ul><hr style="height: 1px;margin: 0;margin-top: 10px;margin-bottom: 10px;border: none;border-top: 1px solid black;"/><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">EXPRESS 2026 期待您的投稿，相聚旧金山湾区，共话 AI 原生时代的软件研运一体化的未来！</span></strong></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link mp_article_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr<a class="wx_topic_link" topic-id="mo30au3f-inkoz4" style="color: #576B95 !important;" data-topic="1" data-recommend="">#qq</a>.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp"><mp-common-profile class="custom_select_card mp_profile_iframe mp_common_widget" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA=="></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://conf.researchr.org/home/splash-issta-2026/express-2026#About">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=65d7e287&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495270%26idx%3D1%26sn%3D3e7673892424a6d556cd051eeb27e368">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Sat, 18 Apr 2026 08:03:00 +0800</pubDate>
    </item>
    <item>
      <title>西安电子科技大学 | SLOT：基于图强化学习的溯源图驱动APT检测方法</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495266&amp;idx=1&amp;sn=29c8c0f6ac509461ca1b2a94bac01ad1</link>
      <description>第一是提出了潜在行为挖掘与相似性感知相结合的思路，用于发现图中的深层关系并削弱伪装邻接节点带来的干扰；第二是将攻击链重构纳入整体框架，不仅判断节点是否异常，还进一步组织出更具阶段性和上下文关联的攻击路径。</description>
      <content:encoded><![CDATA[<p>原创 <span>龙函城</span> <span>2026-04-17 18:30</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=001106a5&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnzjcX4EbjQmVlaOic0LxfPvzsQ6jlBRhjg9JFwpULicDNvav8CUl6Z7KYZ2fyujo93dwcblOXgvs84EPZOic6gj2ECmnAdPwxJaho%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>第一是提出了潜在行为挖掘与相似性感知相结合的思路，用于发现图中的深层关系并削弱伪装邻接节点带来的干扰；第二是将攻击链重构纳入整体框架，不仅判断节点是否异常，还进一步组织出更具阶段性和上下文关联的攻击路径。</p>
  <div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.6555555555555556" data-type="png" data-w="1080" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011611" src="https://wechat2rss.xlab.app/img-proxy/?k=36555dd4&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnyxA8GLO3wgUKJyRnLvG1pzlUhe7rofxqYqB7rTWOtoOlxGjRs9WOPzXadVDYT5g9FHbogq68Kb6DJ6Enj6prhTF8no8g0SBW4%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><blockquote style="border: none;display: block;font-size: 0.9em;overflow: auto;overflow-scrolling: touch;border-left: 3px solid rgba(0, 0, 0, 0.4);background: rgba(0, 0, 0, 0.05);color: #6a737d;padding-top: 10px;padding-bottom: 10px;padding-left: 20px;padding-right: 10px;margin-bottom: 20px;margin-top: 20px;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;color: black;line-height: 26px;text-align: justify;"><em style="font-style: italic;color: black;"><span leaf="">原文标题：SLOT: Provenance-Driven APT Detection through Graph Reinforcement Learning</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">原文作者：Wei Qiao, Yebo Feng, Teng Li, Zhuo Ma, Yulong Shen, Jianfeng Ma, Yang Liu</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">发表会议：CCS&#39;25</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">笔记作者：龙函城</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">主编：黄诚@安全学术圈</span></em></p></blockquote><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">研究概述</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">高级持续性威胁（APT）通常具有潜伏周期长、攻击阶段多、行为隐蔽且伴随伪装等特点，因此传统检测方法很难同时兼顾检测精度、误报控制与分析可解释性。论文指出，现有基于溯源图的APT检测方法主要包括统计型、规范型和学习型三类：统计型方法容易将低频但正常的行为误判为攻击；规范型方法依赖专家知识和规则维护，其适应性不足；而学习型方法虽然能够借助图神经网络建模系统行为，但在攻击者将正常行为嵌入攻击链进行伪装时，仍容易受到邻接节点干扰，导致恶意节点表示失真。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">针对上述问题，本文提出Slot，一种基于图强化学习的溯源图驱动APT检测方法。其核心思路是：首先从系统审计日志中构建溯源图，将进程、文件、网络流等实体及其交互关系表示为图结构，并提取进程名、命令行参数、文件路径、IP 地址、端口等语义信息作为节点初始特征；随后通过潜在行为挖掘机制，在原始系统调用关系之外进一步发现多跳路径中的隐藏关系，以增强对行为因果链、上下文依赖和间接攻击痕迹的刻画能力；在此基础上，引入图强化学习，根据节点的语义与拓扑相似性自适应筛选邻接节点，尽可能过滤伪装性较强的干扰节点，从而提升图表示学习的准确性与鲁棒性。</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.2759259259259259" data-type="png" data-w="1080" style="display:block;margin:0 auto;max-width:100%;width:526px;height:145px;" data-imgfileid="100011610" src="https://wechat2rss.xlab.app/img-proxy/?k=d655794f&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnzN3Q3HvN1f6icLB711kia5DBSd1mHc2f4qn9PPKNEdibueGcO4P5icxZ9QWTX5ickXLV6pwtnk2IygsZaC4GcyVb6wSv2VexQAtIjg%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: center;"><span leaf="">图1 Slot方法整体架构</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如图1所示，Slot的整体流程主要包括五个阶段：图构建、潜在行为挖掘、基于图强化学习的嵌入表示、威胁检测以及攻击链重构。其中，在表示学习阶段，Slot先融合节点属性特征与拓扑特征计算相似性，再通过Bernoulli multi-armed bandit形式的强化学习机制动态调整不同关系下的邻域筛选阈值，引导GNN在多关系图上完成更稳健的聚合。检测阶段则结合MLP与Isolation Forest，一方面识别已学习到的良性与恶意模式，另一方面补充发现未知异常行为；最后，系统进一步结合ATT&amp;CK中的TTP编码与标签传播方法，对离散异常节点进行聚合和重构，生成更紧凑、更易于人工分析的攻击链结果。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">与一般基于溯源图的检测方法相比，Slot的特点在于它并未将图神经网络直接用于异常分类，而是进一步把邻接节点是否可信作为表示学习中的关键问题来处理。为此，论文设计了相似性感知机制，并利用强化学习动态优化不同关系下的邻居保留比例，使模型在聚合上下文信息时尽量保留真正相关的邻接节点及其关系信息，削弱伪装正常行为对恶意节点表征的干扰。同时，Slot只使用少量恶意样本作为监督信号，结合半监督学习完成训练，并在输出层进一步重构攻击链，使系统结果不再停留于单个异常节点，而能够形成更具阶段性和可解释性的攻击路径。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">实验结果表明，Slot在StreamSpot、Unicorn Wget和DARPA E3等多个公开数据集上都取得了较好的检测效果，整体检测准确率约为99%，并在对抗伪装场景下表现出优于现有方法的鲁棒性。论文还指出，借助图强化学习机制，Slot能够更有效地抵御攻击者通过修改图邻域结构来模仿良性行为的对抗策略，同时在案例分析中展示出较好的攻击路径重构能力。Slot不仅提升了APT检测精度，也增强了结果的可解释性。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">贡献分析</span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点1</span></strong><span leaf=""> ：提出APT检测方法Slot</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文提出了Slot，一种基于图强化学习的溯源图驱动APT检测方法。该方法能够在大规模审计日志中有效识别恶意活动，并在检测过程中进一步完成攻击链重构，从而为后续有针对性的防御策略制定提供支持。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点2</span></strong><span leaf=""> ：提出高级图挖掘技术，用于发现图中的多层隐藏关系</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文提出了一种高级图挖掘机制，能够有效挖掘图中的多层隐藏关系，包括因果关系、上下文关系以及间接关联关系等。这使得模型不仅能够关注局部直接连接，还能够从更深层次上理解APT攻击过程中跨节点和跨路径的行为依赖。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点3</span></strong><span leaf=""> ：设计有限标签条件下的半监督学习机制</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">Slot通过高效的标签相似性计算，在仅有少量标签的条件下实现半监督学习。这种设计能够在标签受限的情况下显著提升检测性能与模型鲁棒性。对于真实APT检测场景而言，该机制具有较强现实意义，因为高质量恶意标签通常难以大规模获取。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点4</span></strong><span leaf=""> ：将强化学习引入溯源图分析，提升系统对抗鲁棒性</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文将强化学习引入溯源图表示学习过程，在融合节点语义特征与拓扑特征的基础上，通过强化学习自适应调整邻接节点选择与关系聚合策略，从而减轻伪装正常行为对恶意节点表示的干扰。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点5</span></strong><span leaf=""> ：通过真实数据集实验验证了方法的有效性与实用价值</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文在多个真实APT数据集上对Slot进行了全面评估。实验结果表明，Slot在APT检测准确性、对抗攻击鲁棒性以及攻击链分析支持能力方面均表现突出，验证了其在APT检测与防御支撑中的实际应用潜力。</span></p></div></li></ul><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">实验结果分析</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文围绕半监督检测能力、与现有方法的整体比较、系统开销、模块有效性、对抗鲁棒性以及告警验证效果等方面进行了系统实验。实验数据包括批日志层面的 StreamSpot、Unicorn Wget，以及实体级的 DARPA E3（TRACE、CADETS、THEIA）数据集，既覆盖了相对简单的场景，也包含了具有较强伪装性和更复杂攻击链结构的APT场景。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">表1 Slot方法与当前主流先进方法的对比结果</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.8669590643274854" data-type="png" data-w="684" style="display:block;margin:0 auto;max-width:100%;width:472px;height:409px;" data-imgfileid="100011612" src="https://wechat2rss.xlab.app/img-proxy/?k=441fd58f&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnx4RgxpmUwJFibmRTQSSAhbn18XuEJ7FWaxeuzjNT4ICsLAY9AUA9Viaqq7ibicSYd9CsruGC20GfHqic7MMz7OyiavOTrOvnPjszOfo%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从整体检测效果来看，Slot 在多个数据集上都取得了较优结果。表1显示，在StreamSpot、Unicorn Wget和DARPA E3三组数据上，Slot的Accuracy、Recall和F1-Score整体优于Log2vec、Threatrace、FLASH、MAGIC等方法，同时保持较低的误报率。在较简单的StreamSpot数据集上，多种方法都能取得较好效果，但Slot依然接近最优；而在伪装性更强的Unicorn Wget和DARPA E3数据集上，Slot的优势更加明显，这说明其对复杂APT行为和伪装上下文具有更强的区分能力。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">表 2 批日志级与事件级训练数据集中恶意结构占比对检测性能的影响</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.4006024096385542" data-type="png" data-w="664" style="display:block;margin:0 auto;max-width:100%;width:512px;height:205px;" data-imgfileid="100011608" src="https://wechat2rss.xlab.app/img-proxy/?k=95efa9af&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnwqyebibeA3eH0UliaaQcibeAxCicdtelppbJBFQT0g8C2xt4ibp3DiakgrXk7EZ5D1XGZDoEPAdRFnJ2rzC2M5OhuNYrXjJibNib4QQnE%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">在半监督场景下，Slot 同样表现出较好的有效性。论文仅使用少量恶意结构作为监督信号，通过标签相似性计算和过采样完成训练。表2所示，当训练集中的恶意结构比例从5%提升到10%时，模型性能显著提升，但即使监督信号较少，Slot依然能够获得较高的检测准确率。这说明该方法在标签有限的真实安全场景中具有较好的实用性。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">表 3 在 300K 节点训练图上 Slot 各阶段的时间开销及占比</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.2127659574468085" data-type="png" data-w="705" style="display:block;margin:0 auto;max-width:100%;width:517px;height:110px;" data-imgfileid="100011609" src="https://wechat2rss.xlab.app/img-proxy/?k=1da63b10&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnxdOicYhJdpsReSPfh08N2PVyUMic5ribkDq4SX3WPicgSarlP2yEicQudbuVDd3c3S9x1I2VtIHuJQiaSxLnBD6p3iciaFf75k1gPP7x4%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从运行代价来看，Slot 的主要额外开销来自强化学习过程。表3所示，在300K节点训练图上，图构建耗时46.12秒，强化学习耗时331.94秒，GNN部分耗时64.69秒，其中强化学习占总开销的74.97%。这说明Slot的性能提升并非无代价获得，而是主要通过邻域筛选与鲁棒表示学习换来的。不过，这种开销是可以接受的，因为它带来了更强的对抗鲁棒性和更高的检测质量。</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.5007215007215007" data-type="png" data-w="693" style="display:block;margin:0 auto;max-width:100%;width:479px;height:240px;" data-imgfileid="100011614" src="https://wechat2rss.xlab.app/img-proxy/?k=234ff29d&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnwto9LXYgBvsC1iaziaHeOBu8GGow1l4nh4F8ickrFHw2aFfcj6eIachd6Hy4qcEglUc2nyicJzQ8lr0IOFcZFG3ibzYyrdOd2ZnwwM%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">图 2 对抗攻击鲁棒性结果</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文构造了mimicry attack，通过在攻击图中不断插入良性结构来干扰检测模型。从图2可以看出，随着插入的良性事件数量增加，FLASH的检测得分明显下降，并逐渐接近异常判定阈值；相比之下，Slot的得分始终保持在较高水平，波动也相对更小。该结果表明，Slot通过相似性感知和强化学习筛选邻接节点，能够更有效地削弱伪装良性行为对恶意节点表示的干扰，因此在对抗场景下具有更强鲁棒性。</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.403125" data-type="png" data-w="640" style="display:block;margin:0 auto;max-width:100%;width:509px;height:205px;" data-imgfileid="100011613" src="https://wechat2rss.xlab.app/img-proxy/?k=45f064ae&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnxc4y9OsBo6nqC3z8yYkMMIpTiboup9jRTQ5USHjcoU9MX2hxyElp1UMZIxArqMgJ7zTwZEjFRkCc0H2ib5R8YQca6YzHWzsDeaM%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">图 3 告警数量与攻击链误报控制效果</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">图3所示，论文还评估了Slot在人工分析支持上的价值。在DARPA E3的TRACE、CADETS和THEIA数据集上，原始alert-node数量显著高于最终生成的attack-chain数量，而攻击链内部的precision仍保持在较高水平。这说明Slot在攻击链重构后，能够有效压缩分析人员需要人工检查的对象规模，同时较好地控制误报。其不仅能够检测异常，还能够把离散告警组织成更紧凑、更可解释的攻击路径，从而提升告警验证效率。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文点评</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">总体来看，这篇论文的价值在于，它抓住了APT检测中的一个关键现实问题：在基于溯源图的分析中，攻击者往往会通过嵌入正常行为来伪装攻击路径，从而干扰模型判断。相比传统方法容易受到上下文噪声影响、在复杂场景下鲁棒性不足的问题，Slot进一步将图强化学习引入溯源图分析，使检测过程不仅关注异常识别，也关注邻接关系是否可信，因此整体思路较为清晰，也更贴近真实APT分析场景。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从方法设计上看，论文最突出的亮点有两个。第一是提出了潜在行为挖掘与相似性感知相结合的思路，用于发现图中的深层关系并削弱伪装邻接节点带来的干扰；第二是将攻击链重构纳入整体框架，不仅判断节点是否异常，还进一步组织出更具阶段性和上下文关联的攻击路径。这样一来，系统输出的不再只是零散异常节点，而是更具解释性和分析价值的攻击结果，这一点比单纯的图异常检测更有实际意义。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，这篇论文也存在一定局限。其方法整体链路较长，涉及图构建、隐藏关系挖掘、强化学习筛选、分类检测和攻击链重构等多个环节，因此实现与部署成本相对较高；同时，虽然论文在对抗伪装场景下表现出较好鲁棒性，但对于更强的开放环境攻击和未知行为，仍可能面临误报问题。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文文献</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">Qiao, Wei, et al. &#34;Slot: Provenance-Driven APT detection through graph reinforcement learning.&#34; Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security. 2025.</span></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://dl.acm.org/doi/pdf/10.1145/3719027.3744788">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=2aa3ae88&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495266%26idx%3D1%26sn%3D29c8c0f6ac509461ca1b2a94bac01ad1">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Fri, 17 Apr 2026 18:30:00 +0800</pubDate>
    </item>
    <item>
      <title>行业会议 | 第七届网络空间安全前沿论坛（常安论坛全议程）</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495255&amp;idx=1&amp;sn=88e9c067ec779342ca956f56d96267c3</link>
      <description></description>
      <content:encoded><![CDATA[<p><span>常安论坛</span> <span>2026-04-15 11:31</span> <span style="display: inline-block;">陕西</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=d8960807&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnyd6cibL6tFFic3ApriaVWUWmpSGWTV3c0SAJfAJ10U7qmPDia4WM4J104Y4QMC36SjUWmib3X5MZrkiaNQsy0NqLcP2TibNickiaCSFbiak%2F0%3Fwx_fmt%3Djpeg"/></p>
  
  <p style="text-align: center;" nodeleaf=""><img data-aistatus="1" class="rich_pages wxw-img js_insertlocalimg" data-ratio="17.86586695747001" data-s="300,640" data-type="webp" data-w="917" type="block" data-imgfileid="100011604" src="https://wechat2rss.xlab.app/img-proxy/?k=04dae671&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnxqia8EvIf4L2bVmkXiaPduHvH5zZefhpMNPN5m69VhqvEUVuoUXUs46ic3TTBD6ZffNdOSIPu4rSVPkJf2iboicBfjjOUVhibDic6qWg%2F640%3Fwx_fmt%3Dwebp%26from%3Dappmsg"/></p><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="letter-spacing: 0.578px;border-style: solid;border-right-width: 0px;border-bottom-width: 0px;border-left-width: 0px;border-color: rgba(0, 0, 0, 0.1);transform-origin: 0px 0px;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>



<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=06fd7ef9&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495255%26idx%3D1%26sn%3D88e9c067ec779342ca956f56d96267c3">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Wed, 15 Apr 2026 11:31:00 +0800</pubDate>
    </item>
    <item>
      <title>北京大学 | KnowHow：面向可解释且准确溯源分析的高层CTI知识自动应用方法</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495249&amp;idx=1&amp;sn=d2c557e4914d44e74fb24b743a4f4198</link>
      <description>本文提出KNOWHOW，一种面向在线溯源分析的CTI知识驱动方法，旨在自动将CTI报告中的高层攻击知识应用到底层系统事件检测中。</description>
      <content:encoded><![CDATA[<p>原创 <span>龙函城</span> <span>2026-04-14 18:30</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=16418620&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnxNEKBicnqsrrpUftm5v2s1gL6CLYR55z3csaIzIa0IhTZZQFRzPvowO5xfCJaYSlicphbr4aRaMR39QqsrxGMQWEkLQVsquOic0E%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>本文提出KNOWHOW，一种面向在线溯源分析的CTI知识驱动方法，旨在自动将CTI报告中的高层攻击知识应用到底层系统事件检测中。</p>
  <div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011596" data-ratio="0.30950775455158464" style="display: block;margin: 0 auto;max-width: 100%;" data-type="png" data-w="1483" src="https://wechat2rss.xlab.app/img-proxy/?k=35c49096&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnwExnrL74riaMHicOialJ8k7WGRkvmYHVFWwNaC27j6GC5jKRVINBrqQEj4TohuvECtF3RSwnibblpEpgdEwPpRjNegrlsXtwYVb0w%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><blockquote style="border: none;display: block;font-size: 0.9em;overflow: auto;overflow-scrolling: touch;border-left: 3px solid rgba(0, 0, 0, 0.4);background: rgba(0, 0, 0, 0.05);color: #6a737d;padding-top: 10px;padding-bottom: 10px;padding-left: 20px;padding-right: 10px;margin-bottom: 20px;margin-top: 20px;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;color: black;line-height: 26px;text-align: justify;"><em style="font-style: italic;color: black;"><span leaf="">原文标题：KnowHow: Automatically Applying High-Level CTI Knowledge for Interpretable and Accurate Provenance Analysis</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">原文作者：Yuhan Meng, Shaofei Li, Jiaping Gui, Peng Jiang, Ding Li</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">发表会议：NDSS 2026</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">笔记作者：龙函城</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">主编：黄诚@安全学术圈</span></em></p></blockquote><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">研究概述</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">高级持续性威胁（APT）攻击通常具有潜伏时间长、攻击阶段多、行为链条复杂等特点。这使得仅依赖单点告警、简单规则匹配或纯数据驱动异常检测的方法很难准确还原攻击过程。现有溯源分析方法虽然能够从系统事件中挖掘攻击痕迹，但往往面临两个突出问题：一是误报较多，二是检测结果缺乏可解释性，安全分析人员往往只能看到若干异常节点或异常路径，却难以理解这些结果分别对应什么攻击技术、处于攻击链的哪个阶段。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">另一方面，网络威胁情报（CTI）中包含大量高层攻击知识，例如攻击者使用的技术、操作对象、行为目标以及攻击阶段关系。这些知识天然适合用于APT分析，但问题在于，CTI通常以自然语言形式存在，而实际检测系统处理的却是文件访问、命令执行、网络连接等低层系统事件，两者之间存在明显的语义鸿沟。传统IoC驱动方法大多只关注文件名、IP地址、哈希值等具体指标，对攻击变体和跨环境场景的适应能力较弱；而依赖人工规则的高层知识方法又通常构建成本高、可扩展性不足。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">针对上述问题，本文提出KNOWHOW，一种面向在线溯源分析的CTI知识驱动方法，旨在自动将CTI报告中的高层攻击知识应用到底层系统事件检测中。该方法的核心思想是：首先从CTI报告中抽取可用于检测的高层攻击知识，并构建CTI Knowledge Database（CKD）；随后，通过提出的gIoC（General Indicator of Compromise）表示，将攻击知识抽象为“谁对什么对象执行了什么操作”；最后，再将底层系统事件进行语义提升，与gIoC进行模糊匹配，并结合APT攻击生命周期的阶段逻辑完成攻击推理，从而实现对APT攻击的准确检测与可解释分析。</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011597" data-ratio="0.29074074074074074" style="display:block;margin:0 auto;max-width:100%;width:503px;height:146px;" data-type="png" data-w="1080" src="https://wechat2rss.xlab.app/img-proxy/?k=23578b52&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnweqGqDFc7Sib3FsJwmgrJqa4du7GW7PBRv5fKdcKhicXAjZheQ94IEHvdIiajatQaoVcjrsMN9uKJDnL8aMBbmqETKqicJQicMRbYA%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: center;"><span leaf=""><span textstyle="" style="font-size: 15px;">图 1 KNOWHOW 方法整体架构</span></span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如图1所示，KNOWHOW 的整体流程主要分为三个阶段。首先，系统对输入的系统事件流进行检测，将每个事件与CKD中的攻击技术知识进行匹配；若某事件能够匹配到至少一个ATT&amp;CK技术条目，则被标记为异常事件。其次，系统以这些异常事件为种子，在线构建对应的溯源图，并在扩展过程中引入关联节点，以减少仅做事件级检测可能带来的漏报。最后，系统将候选攻击图映射到APT攻击生命周期模型中，对攻击阶段的时序关系和完整性进行推理，删除不符合攻击逻辑的阶段或节点，最终输出更加精确且具有解释性的攻击告警。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">与传统基于统计异常分数的方法不同，KNOWHOW更强调基于知识的解释性检测。它不是简单判断某个事件是否异常，而是进一步回答：该事件命中了什么攻击技术、属于APT链条中的哪个阶段、为什么这些事件能够共同构成一次可信攻击。这种设计使得系统输出不再只是低层日志片段，而是能够形成带有技术语义和攻击阶段含义的检测结果，从而更贴近真实安全运营场景的分析需求。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">贡献分析</span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点1</span></strong><span leaf=""> ：提出面向APT检测的CTI知识驱动在线溯源分析框架KNOWHOW</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文提出了KNOWHOW，目标将CTI报告中的高层攻击知识自动应用到底层系统事件检测中，实现对APT攻击的在线、准确且可解释分析。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点2</span></strong><span leaf=""> ：提出新的高层攻击知识表示gIoC，并构建CKD知识库</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">为了弥合CTI自然语言描述与系统底层事件之间的语义差距，论文提出了一个新的紧凑知识表示gIoC，用于刻画高层攻击知识。同时构建了新的知识库CKD对这些知识进行统一管理。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点3</span></strong><span leaf=""> ：设计基于APT生命周期阶段的攻击推理方法</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">在完成事件与知识匹配后，论文进一步设计了新的攻击推理方法。该方法基于从CKD查询结果中识别出的ATT&amp;CK技术及其对应的APT攻击生命周期阶段，对候选攻击路径进行推理与筛选。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点4</span></strong><span leaf=""> ：在公开常用数据集和工业数据集上进行了系统评估</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">作者对KNOWHOW在多个广泛使用的数据集以及工业数据集上进行了全面实验。实验结果表明，KNOWHOW在APT检测中能够较好满足准确、有效且可解释的要求。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点5</span></strong><span leaf=""> ：在OpenEuler生态中完成真实安全运营环境部署验证</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">除了实验评估外，KNOWHOW已经部署在OpenEuler生态中进行验证，说明KNOWHOW不仅具有研究价值，也具备一定的工程可落地性。</span></p></div></li></ul><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">代码分析</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">代码链接</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf=""><a href="https://github.com/myh0301/KNOWHOW" target="_blank">https://github.com/myh0301/KNOWHOW</a></span></p></li></ul><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">使用类库分析</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">该项目整体采用Python实现，属于典型的安全研究原型代码。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">代码实现难度与工作量评估</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从实现角度看，该项目难度属于中等偏上。其难点在于整体流程较长与模块间依赖较强。代码不仅要完成CTI文本中的知识抽取与表示构建，还要对底层系统日志进行语义增强和技术打标，并进一步完成告警图构建、生命周期映射与结果输出。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码关键实现功能</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（1）CTI知识抽取模块。该模块的主要功能是从CTI文本中抽取结构化攻击知识，并转化为论文中的gIoC表示。gIoC 采用 &lt;subject, verb, object&gt; 三元组结构来表示攻击知识，这一模块承担了把自然语言形式的攻击描述转化为可计算知识表示的任务。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（2）知识嵌入与语义对齐模块。在抽取得到gIoC之后，系统会进一步对这些知识进行语义编码，并把它们组织为可供查询和匹配的知识表示。其核心思想是把CTI知识与系统日志都映射到统一语义空间中，再通过相似度比较判断日志中是否存在威胁行为。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（3）系统日志语义增强与异常事件检测模块。该模块面向底层系统日志，对原始事件进行语义增强和标准化处理，再结合前面得到的知识表示进行比对，从而识别潜在攻击行为。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（4）溯源图构建与事件级告警生成模块。在完成单条日志事件检测后，系统会进一步将相关事件组织成图结构，生成事件级告警和候选攻击链。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（5）APT生命周期推理与告警去噪模块。在图构建基础上，系统还会结合APT攻击生命周期对候选告警进行推理、筛选和重组，以验证其是否符合攻击链条的阶段逻辑。该模块进一步提升结果的可解释性和可信度，使最终告警能够以更接近真实攻击过程的形式呈现，并支持后续攻击报告生成。</span></p></div></li></ol><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文点评</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">总体来看，这篇论文的价值在于它抓住了一个非常实际的问题：如何把CTI报告中的高层攻击知识真正应用到底层系统事件检测中。相比传统数据驱动方法容易产生误报、结果难以解释的问题，KNOWHOW把自然语言形式的攻击知识转化为可用于实际检测的结构化信息，因此整体思路比较清晰，也更贴近真实安全分析场景。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从方法设计上看，论文最突出的亮点有两个。第一是提出了gIoC这种中间表示，用来描述攻击主体、动作和目标，从而增强了CTI知识的可计算性和泛化能力；第二是引入APT攻击生命周期进行攻击推理，不仅判断事件是否可疑，还进一步分析其是否符合攻击链条的阶段逻辑。这样一来，系统输出的不再只是零散异常事件，而是更具上下文和解释性的攻击结果，这一点比单纯的异常检测更有实际意义。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，这篇论文也存在一定局限。其方法本质上仍然依赖已有CTI知识库的覆盖范围，因此面对与现有知识差异较大的全新攻击时，检测能力可能会受到影响；同时，真实环境中的高风险合法行为与攻击行为之间往往边界模糊，这也使系统在开放场景下仍可能出现误报。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文文献</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">Meng, et al. “KnowHow: Automatically Applying High-Level CTI Knowledge for Interpretable and Accurate Provenance Analysis.” Network and Distributed System Security Symposium (NDSS 2026). 2026.</span></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://arxiv.org/pdf/2509.05698">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=1dad2f57&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495249%26idx%3D1%26sn%3Dd2c557e4914d44e74fb24b743a4f4198">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Tue, 14 Apr 2026 18:30:00 +0800</pubDate>
    </item>
    <item>
      <title>四川大学；清华大学等：黑客还是幻觉？基于大语言模型的自动化渗透测试全面分析</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495243&amp;idx=1&amp;sn=d863eee8a987d434f3a2f3a2b5b608be</link>
      <description>本研究首次针对LLM驱动的AutoPT框架进行知识系统化分析，并开展全面而深入的大规模实证评估，旨在厘清当前技术真实的能力边界，并为后续研究提供参考。</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-12 11:47</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=5c1a5c16&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnyHptFRICgvAu1Htor2ziaqkVibgXBNAvdcxSFe5yJxmQNBaDG0FoGmazjXsGf6Ixt3tQicIicI4XbGDaWInMWnchWnt7jBroV2jVE%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>本研究首次针对LLM驱动的AutoPT框架进行知识系统化分析，并开展全面而深入的大规模实证评估，旨在厘清当前技术真实的能力边界，并为后续研究提供参考。</p>
  <div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0px 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;overflow-wrap: break-word;text-align: justify;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#34;PingFang SC&#34;, Cambria, Cochin, Georgia, Times, &#34;Times New Roman&#34;, serif;" data-pm-slice="0 0 []"><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.7923433874709976" data-type="png" data-w="862" style="display: block;margin: 0px auto;max-width: 100%;width: 543px;height: 430px;" data-imgfileid="100011589" src="https://wechat2rss.xlab.app/img-proxy/?k=5a3f737c&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnzve77DtPwoeeekTXJX5tfpznlcwuhFC2jmZuQhUVw6aZngak151lkfUgP7wBoulfG0tiaiaicuBqxVDnoQw3VQU01Pr2mwM5WKCs%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><blockquote style="border-top: none;border-right: none;border-bottom: none;border-image: initial;display: block;font-size: 0.9em;overflow: auto;border-left: 3px solid rgba(0, 0, 0, 0.4);background: rgba(0, 0, 0, 0.05);color: rgb(106, 115, 125);padding: 10px 10px 10px 20px;margin-bottom: 20px;margin-top: 20px;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;color: black;line-height: 1.5em;text-align: left;"><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">原文标题：Hackers or Hallucinators? A Comprehensive Analysis of LLM-Based Automated Penetration Testing</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">原文作者：  Jiaren Peng, Zeqing Li, Chang You, Yan Wang, Hanlin Sun, Xuan Tian, Shuqiao Zhang, Junyi Liu, Jianguo Zhao, Renyang Liu, Haoran Ou, Yuqiang Sun, Jiancheng Zhang, Yutong Jiao, Kunshu Song, Chao Zhang, Fan Shi, Hongda Sun, Rui Yan, Cheng Huang</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">原文链接：原文链接：<a href="https://arxiv.org/abs/2604.05719" target="_blank">https://arxiv.org/abs/2604.05719</a></span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">发表会议：arXiv</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">笔记作者：孙汉林，彭佳仁@安全学术圈</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 14px;">主编：黄诚@安全学术圈</span></span></em></p></blockquote><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">1. 研究背景</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;"><span leaf="">随着大语言模型（Large Language Model, LLM）的飞速发展，自动化渗透测试（Automated Penetration Testing, AutoPT）迎来了前所未有的机遇，进而催生出一系列旨在实现端到端自主攻击的研究框架。然而，在相关研究呈现井喷之势的背后，现有的AutoPT研究普遍缺乏系统性的架构分析，也鲜有在统一基准下进行的大规模实证比较。为此，本研究首次针对LLM驱动的AutoPT框架进行知识系统化分析，并开展全面而深入的大规模实证评估，旨在厘清当前技术真实的能力边界，并为后续研究提供参考。</span></p><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011591" data-ratio="1.24093567251462" style="display: block;margin: 0px auto;max-width: 100%;" data-type="png" data-w="855" src="https://wechat2rss.xlab.app/img-proxy/?k=d91c157b&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnxkvIa6bFTHfsNShsDiagR127v11SvltpOm1oJR7gRcOq62icu8BpOb9dTvbejVdCH1c8zq45v9DLXpSuUzMHdmOxR6GEE2R3Qks%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;"><span leaf="">2 实验设计</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">在系统化体系分析层面，本研究构建了一套多维度的分类分析框架，围绕协作范式、任务规划、智能体记忆、工具执行、外部知识与基准测试六大核心维度，对现有AutoPT框架进行了系统性解构。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">在实证层面，研究团队在统一实验设置下，本研究对比了13个具有代表性的开源AutoPT框架以及2个基于成熟的商业AI编程助手构建的基线框架。整个实验过程累计消耗超100亿Token，生成超1500份执行日志，并由15位以上具备网络安全专业背景的研究人员，历时四个月，对上述自动化渗透测试日志进行了系统性人工审查与框架机制剖析，以确保评估结果的可靠性与分析结论的严谨性。</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">3 实验结果</span></h3><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011590" data-ratio="1.3048065650644782" style="display: block;margin: 0px auto;max-width: 100%;" data-type="png" data-w="853" src="https://wechat2rss.xlab.app/img-proxy/?k=64f16bdf&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnwjYToHPInvXjYav7EcFX0CHOwt5ibcHrpd8mZoQwCcwYZwILjiceMhJY87fGrNzwIa4QvP1Sx6Kv4P5msVQXPicwMtgrca5iajL4Y%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">表中圆圈符号表示模型在各靶场挑战中是否成功获取Flag，全黑表示两次均成功；全白表示两次均失败；左黑右白表示V1成功、V2失败；左白右黑表示V1失败、V2成功。“S”列代表框架的总得分，“E”列代表框架的在所有easy靶场上的总得分，“M”列代表框架的在所有Medium靶场上的总得分，“H”列代表框架的在所有Hard靶场上的总得分，每次成功获取Flag将根据靶场难度进行计分：Easy挑战得2分，Medium挑战得3分，Hard挑战得5分。</span></p><figure data-tool="mdnice编辑器" style="margin: 10px 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011588" data-ratio="0.8591715976331361" style="display: block;margin: 0px auto;max-width: 100%;" data-type="png" data-w="845" src="https://wechat2rss.xlab.app/img-proxy/?k=dc303d22&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnwCzAYeejphOdft0osIDJ3MgKKr2icQsib3339XbGic1h1QiaBDRVFtw7bYrxD28KCkflYian3au9CZgO1WXFaPZUsxRx0kBLJ4uH0Y%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">基于以上大规模实验数据，本研究揭示了若干关键发现：</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(1) 单智能体架构在多数中低难度任务上的综合表现并不逊色甚至超越了复杂的多智能体框架；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(2) 单智能体架构在面对复杂问题上的单次调用的Token消耗高于大多数多智能体框架；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(3) 记忆管理是影响当前AutoPT框架性能的关键因素；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(4) 外部知识库在多数情况下未能带来正向增益，不匹配的检索结果极易误导智能体陷入错误的探索方向，传统RAG难以带来有效表现；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(5) 工具池的规模扩张并不等价于渗透能力的提升；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(6) 工具受限时，框架会触发以Python执行等的补偿机制，但该机制在困难渗透测试场景中效果有限；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(7) AI编程智能体仅凭基础的工具和简单的提示词便能展现出惊人的竞争力；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(8) 不同LLM基座的表现不同，部分LLM表现出的特点需要框架与其进行适配；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(9) 公开CVE漏洞的稳定利用依赖动态维护的针对性知识库；</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">(10) 幻觉现象普遍存在，尤其是CTF场景下的Flag幻觉。</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">4 总结</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">本文通过系统化的架构分析与大规模实证，对LLM在AutoPT中的实际能力与现有局限进行了全面评估，填补了AutoPT领域在严谨实证支持与定量比较方面的研究空白，为未来框架的设计范式提供了方向性的指引。研究结果表明，在利用大型模型推动自动化攻防的过程中，需保持审慎的工程评估视角。若缺乏完善的底层记忆管理机制与动态场景适配能力，盲目扩展多智能体协作规模或大量集成外部知识，极易孕育出“Hallucinators”而非真正的“Hackers”。本文为网络安全领域研究者提供了结构化的评估分类体系及可持续更新的验证基准，为LLM时代下更智能、更可靠的自动化渗透测试技术发展贡献了相应的学术基础。</span></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://arxiv.org/abs/2604.05719">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=ee060c88&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495243%26idx%3D1%26sn%3Dd863eee8a987d434f3a2f3a2b5b608be">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Sun, 12 Apr 2026 11:47:00 +0800</pubDate>
    </item>
    <item>
      <title>行业会议 | 2026常安论坛（早鸟票倒计时一周）</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495235&amp;idx=1&amp;sn=d3fb9940df446fa59954b0d4a9434f56</link>
      <description>常安论坛 | 早鸟票倒计时一周！</description>
      <content:encoded><![CDATA[<p><span>常安论坛</span> <span>2026-04-10 22:14</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=d11de5c2&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnz452mTymd8ianJ3YGGWLiafZiaxJu8QAwud2Sf7ibokwdZib6TAjSeS3ypewHOIMUSR3PCFfCbzjkjjib5WhQIroEcCxZKUGtZV1QKw%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>常安论坛 | 早鸟票倒计时一周！</p>
  <p style="text-align: center;margin-bottom: 0px;"><span leaf=""><img class="rich_pages wxw-img js_insertlocalimg" data-aistatus="1" data-imgfileid="100011580" data-ratio="16.157" data-s="300,640" type="block" data-type="webp" data-w="1000" src="https://wechat2rss.xlab.app/img-proxy/?k=71e652dc&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnwLwpxqSh7evwtHMwXHrwavnHB3ibcJ6gLDhE4b9ywwX5bYnsRHcgB0a6yBkBkdO8VXIHoGAgCakm2sJvjBRVvSXgoBHCSoX4g0%2F640%3Fwx_fmt%3Dwebp%26from%3Dappmsg"/></span></p><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="letter-spacing: 0.578px;border-style: solid;border-right-width: 0px;border-bottom-width: 0px;border-left-width: 0px;border-color: rgba(0, 0, 0, 0.1);transform-origin: 0px 0px;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>



<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=d8441cf6&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495235%26idx%3D1%26sn%3Dd3fb9940df446fa59954b0d4a9434f56">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Fri, 10 Apr 2026 22:14:00 +0800</pubDate>
    </item>
    <item>
      <title>大湾区大学 | 网络与信息安全研究中心招聘博士后</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495227&amp;idx=1&amp;sn=7ed11119a9ed6fd7317820e0783910d6</link>
      <description>大湾区大学 | 网络与信息安全研究中心招聘博士后</description>
      <content:encoded><![CDATA[<p><span>安全学术圈</span> <span>2026-04-09 21:31</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=2c7e92a8&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnw7dbtalOtzDDsxIn9rG8uQFkZ4EgQLdgIKe0DtZhSAyw8BUBibDXu5nqLSp4uR8I2zUHWTuJ5uxggeF75wrLRoic2o4g7vFfbEM%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>大湾区大学 | 网络与信息安全研究中心招聘博士后</p>
  <p style="margin-bottom: 0px;letter-spacing: 0.578px;text-wrap: wrap;text-align: left;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011576" data-ratio="0.4222222222222222" data-s="300,640" type="block" data-type="png" data-w="1395" src="https://wechat2rss.xlab.app/img-proxy/?k=c447bf9c&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnweeLsdwyTjwDIuFbZKFjQ1r8DwsT6k6aJRtNOXQWPo1tyA8AOkOHI49kEHSee7DwUVVmERfrZ1JHxyj6roNRHyobgSheZO114%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">PART 01 课题组介绍</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">东莞市大湾区高等研究院网络与信息安全研究中心依托数学、量子计算、计算机、微电子等学科，聚焦服务国家战略需求的核心目标，开展高水平学术研究与实际应用，致力于密码学、网络空间安全、抗量子密码、隐私计算的理论与关键技术突破，重点推进在工控安全、人工智能安全、大数据安全方向的实际应用。预期形成兼具理论高度与应用价值的研究成果。服务于大湾区的人才培养与网络空间安全建设，为国家网络空间安全体系建设贡献力量。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">网络与信息安全研究中心主任徐茂智，国家二级教授、博士生导师，曾任北大数学学院教授，北大网络空间安全研究院院长，是中国密码学会首批会士、全国密码专业学位研究生教育指导委员会委员、中国密码学会副理事长、中国网络空间安全学会副监事长、中国密码学会应用工作委员会主任、国家重点研发计划项目首席。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">徐茂智教授主要研究兴趣是密码学及信息安全。其在群表示论中的一项成果被国际著名数学家J. Olsson和J. Brundan等称为“徐氏构造法”在国际数学界享有盛誉，受到1998年国际数学家大会邀请；提出无限域上椭圆曲线密码新理论，对椭圆曲线密码理论和实现做出重要贡献，破译了PKCY等多个重要密码体制，完成多个密码算法设计，并得到重要应用。发表论文近200篇，出版专著4部，多次主持国家重点研发计划项目、国家自然科学基金重大研究计划重点课题、国家973课题、国家863课题、国家自然科学基金面上项目等研究，并取得了一系列突破性的成果。4次获得部级科技进步奖二等奖。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;text-align: justify;"><span leaf="">PART 02 联合招收导师介绍</span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">导师一</span></strong><span leaf="">：林璟锵 教授  中国科学技术大学教授、博士生导师、网络空间安全学院执行院长。长期致力于密码技术在计算机网络系统的安全应用和工程实施。研究成果获美国发明专利5项、中国发明专利30多项，高性能密码算法实现、密钥安全方案、PKI数字证书服务安全等成果获国家科技进步二等奖、党政密码科技进步一等奖、北京市科技进步二等奖等，已广泛应用于我国各种电子签名、电子合同、电子证照、视频加解密等系统。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">导师二</span></strong><span leaf="">：胡红钢 教授  中国科学技术大学教授、博士生导师、网络信息安全系（筹）执行主任。长期从事编码、密码与网络安全的研究，证明了国际上两个重要的猜想，在国内外权威期刊和权威会议上发表30多篇论文，先后主持了国家自然科学基金等项目。其团队2024年在TU Darmstadt格挑战中求解95维LWE问题，刷新国际纪录。此外，还担任《密码学报》领域编委和《信息网络安全》编委，2013年获中国密码学会优秀青年奖，2015年获国家优秀青年科学基金，2017年获中国科学院优秀百人计划奖励，2024年获省级教学成果特等奖。</span></p></div></li></ul><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">PART 03 招聘信息</span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">招聘方向：</span></strong><span leaf=""> 人工智能与安全、密码与信息安全、量子信息与量子密码</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">招聘岗位及职责：</span></strong><span leaf=""> 博士后2名（分属两位导师，可自主选择导师方向），具体职责如下：</span></p></div></li></ul><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">自主开展学术科研，撰写学术论文。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">独立或者协助申请各类科研项目。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">积极拓展横向课题，组织与参加学术交流活动。</span></p></div></li></ol><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">应聘条件：</span></strong></p></li></ul><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">获得相关方向博士学位不超过3年或者即将获得博士学位。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">年龄不超过35周岁。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">至少发表（含接收）2篇SCI论文。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">具备良好的团队合作精神和沟通能力，拥有独立思考和解决问题的能力，能够自主开展研究工作。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">遵守学术道德和单位的规章制度，无违反学术诚信行为记录。</span></p></div></li></ol><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">薪酬待遇：</span></strong></p></li></ul><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">聘期2年。博士后综合年薪35-45万元（含地方政府资助），根据科研工作业绩表现，课题组提供相应的科研绩效奖励；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">除综合年薪外，学校缴纳社会保险和住房公积金；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">支持博士后作为负责人申请国家自然科学基金、中国博士后科学基金以及省、市各级科研项目；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">入选卓越博士后的可聘为副研究员；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">支持博士后申报东莞市有关人才项目，享受购房补贴、租房补贴、生活补助、科技成果转化贡献奖励、医疗保障、境外人才个人所得税优惠等；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">享受带薪年假、健康体检和校内餐厅等福利待遇，学校按属地政策协助解决子女入园入学；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">对于优秀的出站博士后积极推荐申请校内相关岗位。</span></p></div></li></ol><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><strong style="font-weight: bold;color: black;"><span leaf="">应聘方式：</span></strong><span leaf=""> 请有意者提供以下申请材料，并将电子版发送至：miscenter@gbu.edu.cn，邮件主题为 “博士后应聘材料+姓名+意向导师”（意向导师填写“林璟锵教授”或“胡红钢教授”），并请推荐人将推荐信直接发送至上述邮箱：</span></p></li></ul><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">个人简历：含教育和工作经历、科研活动及成果等；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">学历、学位证明文件；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">代表性研究成果证明材料：如论文、专利、科研项目、奖励情况等；</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">相关领域专家推荐信2-3封（含博士生导师推荐信）。</span></p></div></li></ol></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>



<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=dc56c276&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495227%26idx%3D1%26sn%3D7ed11119a9ed6fd7317820e0783910d6">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Thu, 09 Apr 2026 21:31:00 +0800</pubDate>
    </item>
    <item>
      <title>2026年计算机软件新技术全国重点实验室开放课题</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495212&amp;idx=1&amp;sn=521aada569851f048c3b63b67ff17cc9</link>
      <description>2026年4月30日前将申请表寄回南京大学计算机软件新技术全国重点实验室</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-08 09:01</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=69481eb0&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnzqKTyX9yW7kQ2BR6W58JDrXibxBpjOFfCG8v2DSAYvvSXe6IBdCz5whCMBsb8zaq1wZmHcULD6q32r1Nv0IQjgDmebgnDJ9Ndg%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>2026年4月30日前将申请表寄回南京大学计算机软件新技术全国重点实验室</p>
  <p style="margin-bottom: 0px;letter-spacing: 0.578px;text-align: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.4255555555555556" data-s="300,640" data-type="png" data-w="900" style="height: 246px;width: 578px;" data-cropselx1="0" data-cropselx2="578" data-cropsely1="0" data-cropsely2="321" data-imgfileid="100009490" src="https://wechat2rss.xlab.app/img-proxy/?k=e1d3dd2d&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F6Dibw6L070WHqFxBGaQCzyTLqUunVD0DHiaxngHHyv1M90jeVsuXReYSw10gicjKE9c2qN0y3gtsicYgkeKeAW2aGg%2F640%3Fwx_fmt%3Djpeg%26from%3Dappmsg"/></span></p><div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">计算机软件新技术国家重点实验室是在南京大学计算机科学与技术系和计算机软件研究所多年进行软件研究的基础上于</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">1986年底通过专家论证，1987年开始筹建，1990年建成</span></code><span leaf="">，并通过国家验收，被批准向国内外开放。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">计算机软件新技术全国重点实验室以基础研究为支撑，定位于</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">应用基础研究</span></code><span leaf="">，以数字经济与数字中国国家重大战略中“</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">提升关键软件技术创新和供给能力</span></code><span leaf="">”为使命任务导向，致力于科技创新引领产业创新，坚持用科技自主自强支撑产业自主可控。按照《计算机软件新技术全国重点实验室国内开放课题基金管理办法及实施细则》相关规定，现组织申报2026年度国内开放课题，诚邀高等院校、计算机与人工智能领域重点实验室联盟成员、相关科研机构和事业单位的有关科研人员积极申报。相关事宜通知如下：</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">一、目标导向类课题（具体名称自拟；拟资助2-3项）</span></h3><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">可信可解释可成长智能化软件的理论、方法与应用</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">学件的理论、方法与应用</span></p></li></ol><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">二、自由探索类课题</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">1. 新型程序设计与软件方法学</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">软件方法和技术必须一方面适应软件平台的发展，另一方面为新型计算机应用模式提供良好支撑。本方向立项的目标是利用人机物三元融合万物智能互联的新计算场景平台和大语言模型快速发展带来的技术机遇和挑战，突破传统的程序设计风格，以智能化基础软件技术、软件智能体技术、神经符号融合软件技术等为核心，进行新一代软件概念、范式、理论、方法、技术、工具和应用模式的探讨，开拓新的研究方向和新的增长点，为原创性软件新技术的研究与开发奠定基础。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">选题范围：</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">1)面向人机物融合的智能化、网构化软件方法学</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">2)自然语言-形式语言协同的新型程序设计</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">3)神经符号融合软件的构造与可信保障方法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">4)软件智能体系统基础理论与构造方法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">5)分布式、智能化基础软件系统与平台</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">6)软件自适应、动态演化与持续成长方法与技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">7)可信化、智能化的软件质量综合保障技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">8)大语言模型支撑的软件开发与质量保障方法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">2. 机器学习与智能化系统</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">针对软件智能化重要趋势，面向智能数据分析处理的广泛需求，瞄准国际前沿，对机器学习与智能化信息处理的基础理论、关键技术和先进应用进行研究，开拓新的研究方向和新的增长点，为下一代软件技术的研究与开发奠定基础。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">选题范围：</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">1)机器学习理论基础与前沿方法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">2)新型机器学习的计算学习理论</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">3)多源异构数据挖掘理论与技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">4)多模态模式识别方法与技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">5)神经网络新型架构与优化方法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">6)大规模演化计算理论与方法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">7)可成长异构大模型及模型轻量化技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">8)可信知识表示与推理技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">9)大规模符号数值混合优化技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">10)可协同分布式人工智能技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">11)动态自适应智能规划与决策技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">12)智能系统关键技术及应用</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">3. 软件可信保障与自动化工具</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">人工智能赋能社会经济各行各业，人工智能模型和框架支撑软件工程各类研发活动，提升软件研发效率，也带来了可信保障的挑战，而且，人工智能模型嵌入各类软件形成全新智能软件，如智能驾驶系统、机器人系统，复杂智能软件的质量不仅仅由实现软件的代码和代码运行的环境所决定，还依赖于嵌入软件系统的人工智能模型和软件的外部输入数据，需要研究复杂智能软件质量保障与可信性评估技术，以及相应的自动化支撑工具，支持实际应用需要。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">选题范围：</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">1)代码大模型的可信、可解释研究</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">2)基于大模型的代码生成与可信保障</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">3)智能体构造、演化与可信保障研究</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">4)大模型辅助的软件分析、测试、验证技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">5)人工智能系统和框架的质量保证</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">6)复杂关键软件系统智能化可信构造与保障方法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">4.分布式泛在计算与系统安全技术</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">针对面向三元融合新世界关键场景的智能化基础软件，探讨作为基础支撑的分布式计算与系统安全相关的新模式、新方法与新技术，研究支撑人机物三元融合的场景化建模与动态适配、分布式计算架构与系统安全保障的关键技术与平台，通过关键技术研究与平台构建，为智能化应用提供坚实的底层技术支撑，推动人机物三元融合的创新发展。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">选题范围：</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">1)场景化建模与智能感知</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">2)多模态工业具身智能</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">3)感传算控一体化系统</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">4)云边端协同计算架构与机制</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">5)边缘智能计算理论与方法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">6)复杂系统智能调度与优化算法</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">7)分布式机器学习理论与系统</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">8)大模型高效推理与性能优化</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">9)新型存储体系结构关键技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">10)隐私保护与数据安全理论技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">11)智能安全监测与防御关键技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">12)算电碳协同分布式智能计算</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">5. 大数据媒体计算与内容处理平台</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">针对人工智能与多媒体技术相关的理论、方法和应用，研究对学科发展和技术进步有重要影响的新理论和新方法，或对国民经济建设有重大影响并具有广泛应用前景的新技术，以及在相关行业构造示范应用系统并加以推广应用。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">选题范围：</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">1）多模态大模型技术研究及平台构建</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">2）通用视觉理解基础模型关键技术研究</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">3）面向具身场景的智能感知与决策技术</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">4）大模型驱动的自动化知识工程</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">5）大规模复杂场景的智能建模和绘制</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">6）面向大规模数据的基础模型与分布式训练</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">7）智能医药大模型关键技术研究与应用</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">8）大数据驱动的科学发现与复杂系统模拟</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">三、资助对象与申请条件</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">凡研究方向符合开放课题申请指南，具有一定研究经历，</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">中级职称以上或取得博士学位</span></code><span leaf="">的研究人员，均可根据课题指南提出课题申请，经本实验室学术委员会按择优支持的原则，经评审后批准开放课题并获得课题基金资助。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">本实验室每年约可资助</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">开放课题项目40个</span></code><span leaf="">左右，目标导向类课题原则上资助强度</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">不超过4万元/项</span></code><span leaf="">，自由探索类开放课题原则上资助强度</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">不超过2万元/项</span></code><span leaf="">。资助强度将根据每年科技部下达经费情况、开放课题申请数量及水平进行调整。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">根据财政部、科技部财务政策要求，开放课题经费</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">只能外拨至高等院校、科研院所及相关事业单位</span></code><span leaf="">；</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">企业重点实验室人员</span></code><span leaf="">需在我室联合一名固定人员共同申报，经费需在南京大学进行报销使用。</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">四、申报要求和程序</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">申请人从实验室网页（<span textstyle="" style="font-style: italic;text-decoration: underline;"><a href="https://keysoftlab.nju.edu.cn/1611/list.htm" target="_blank">https://keysoftlab.nju.edu.cn/1611/list.htm</a></span>）下载并据实填写开放课题基金申请表，由所在单位审批同意、领导签署意见并加盖公章后，请于</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">2026年4月30日前</span></code><span leaf="">将申请表寄回南京大学计算机软件新技术全国重点实验室（电子版1份、纸质版2份）。</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"></ul><ul style="box-sizing: border-box;margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: rgb(0, 0, 0);list-style-type: disc;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#34;PingFang SC&#34;, Cambria, Cochin, Georgia, Times, &#34;Times New Roman&#34;, serif;font-size: 16px;font-style: normal;font-variant-ligatures: normal;font-variant-caps: normal;font-weight: 400;letter-spacing: normal;orphans: 2;text-align: left;text-indent: 0px;text-transform: none;widows: 2;word-spacing: 0px;-webkit-text-stroke-width: 0px;white-space: normal;background-color: rgb(255, 255, 255);text-decoration-thickness: initial;text-decoration-style: initial;text-decoration-color: initial;" class="list-paddingleft-1"><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">联 系 人：徐烜</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">通信地址：南京市栖霞区仙林大道163号南京大学仙林校区计算机科学技术楼519室</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">邮 编：210093</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">电 话：+86（025）89683467 18912990005</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">传 真：+86（025）89686596</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">E - mail：keysoftlab@nju.edu.cn</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">网 页：<span textstyle="" style="font-style: italic;text-decoration: underline;"><a href="http://keysoftlab.nju.edu.cn" target="_blank">http://keysoftlab.nju.edu.cn</a></span></span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">本实验室热忱欢迎和邀请有关领域的科研人员申请与上述研究方向相关的研究课题，特别欢迎国内外博士后人员来实验室工作，共同推进计算机软件新技术的发展。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf=""><span textstyle="" style="font-style: italic;text-decoration: underline;"><a href="https://keysoftlab.nju.edu.cn/1573/list.htm" target="_blank">https://keysoftlab.nju.edu.cn/1573/list.htm</a></span></span></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="white-space: normal; max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__43" style="margin-bottom: 0px;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);color: rgba(0, 0, 0, 0.5);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a>,</span><span leaf=""><br/></span><strong><span leaf="">有兴趣加入学术圈的请联系secdr#qq.com</span></strong></p></blockquote></p><hr style="letter-spacing: 0.578px;border-style: solid;border-right-width: 0px;border-bottom-width: 0px;border-left-width: 0px;border-color: rgba(0, 0, 0, 0.1);transform-origin: 0px 0px;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://keysoftlab.nju.edu.cn/1573/list.htm">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=e7704538&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495212%26idx%3D1%26sn%3D521aada569851f048c3b63b67ff17cc9">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Wed, 08 Apr 2026 09:01:00 +0800</pubDate>
    </item>
    <item>
      <title>四川大学 | 网络空间安全学院2026年人才招聘</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495216&amp;idx=1&amp;sn=bcd7ba526b02d23900f886b19f806b6a</link>
      <description>其中双百截至日期：2026年4月24日</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-07 09:03</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=afce6af5&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnweGdgnxgAQVX1jIHrAeI4Yqk4GCXytQjysWG0LXLdxxYVFbHFIvVicD7SDpOWIMhbaqXvZruHtfvyLicRSSBMlwNFxBNBAMtbg0%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>其中双百截至日期：2026年4月24日</p>
  <p style="text-align: center;margin-bottom: 0px;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.4255555555555556" data-s="300,640" data-type="png" data-w="900" style="width: 578px;height: 246px;" data-cropselx1="0" data-cropselx2="578" data-cropsely1="0" data-cropsely2="321" data-fileid="100003011" src="https://wechat2rss.xlab.app/img-proxy/?k=c236fd04&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F6Dibw6L070WEF0pjW7A9fxnIQUibYyOox1XY5C69NqeK8WWakCbia0VibXCG4mMR5Oy1m5FGia8dMLEqrw5PXhFv9Rw%2F640%3Fwx_fmt%3Dpng"/></span></p><div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">四川大学自</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">上世纪90年代中期即开展网络与信息安全专业</span></code><span leaf="">的人才培养与科学研究工作。2015年，成立网络空间安全研究院。2016年成立四川大学网络空间安全学院，是国务院学位委员会批准的首批29个网络空间安全一级学科博士点培养单位之一。2019年获得首批网络空间安全博士后科研流动站。2020年网络空间安全专业获批国家一流本科专业建设点。学院拥有良好的教学科研平台和完备的人才培养体系，拥有国家网络空间安全人才培养基地；建有网络空间安全研究院、计算机网络与安全研究所、网络与可信计算研究所和信息安全研究所等特色研究机构。此外，学院建有</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">数据安全防护与智能治理教育部重点实验室、四川省网络靶场工程研究中心、国家智能社会治理实验基地（教育）、四川省信息安全重点实验室、中国信息安全测评技术实验室、四川省大数据分析协同创新中心，以及四川省公安厅-四川大学网络违法犯罪生态研究中心</span></code><span leaf="">等多个省部级实验室与研究中心。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">学院在基础理论研究、技术应用和产业化方面都取得了显著的成绩。先后承担</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">国家重点研发计划项目、国家科技攻关项目、“973”项目、“863”项目以及国家自然科学基金重点项目</span></code><span leaf="">等国家级项目百余项，累计科研经费超过亿元。发表SCI/EI收录学术论文300多篇，牵头参与制定国家标准10余项，授权发明专利逾百项。科研团队面向国家、社会重大需求，研究成果得到广泛应用，取得显著经济和重要社会效益。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">2026年学院拟招聘</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">特批研究员1名、特聘副研究员3名、博士后4名</span></code><span leaf="">，岗位详细要求如下：</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">一、招聘岗位</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">岗位1：特聘研究员</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">聘用方式：年薪制</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">专业：网络空间安全及相关专业和方向</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">应聘条件：博士学位</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">招聘1人 | 学校经费聘用</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">岗位职责:</span></strong></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成特聘研究员岗位要求的科研工作任务，主持国家级科研项目，参与科研奖励申报，产出高水平科研成果，积极进行成果转化；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成特聘研究员岗位要求的教学工作任务，承担本科生课堂教学任务、指导本科生毕业设计等；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">积极参与学科平台建设，积极参加各种学术活动和国际交流；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成校院两级公共服务及学院交办的其他工作。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成岗位聘任合同中规定的其他工作。</span></p></li></ol><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">招聘条件:</span></strong></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">具有良好的思想道德素质和职业道德，品行端正，爱岗敬业，责任心强，具有较强的团结协作精神；身心健康，年龄</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">不超过40岁</span></code><span leaf="">；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">从事网络空间安全领域相关专业或方向，具有国内外高水平大学博士学位；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">有网络空间安全及相关领域工作经历，具备带领团队协同工作的能力；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">具有网络空间安全领域较高学术水平，学术思想活跃，创新能力强，已在本学科领域取得一定的成就，发展潜力大，原则上应达到申报</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">我校研究员专业技术职务成果</span></code><span leaf="">要求。</span></p></li></ol><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">岗位2：特聘副研究员</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">聘用方式：年薪制</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">专业：网络空间安全及相关专业和方向</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">应聘条件：博士学位</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">招聘3人 | 学校经费聘用</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">岗位职责:</span></strong></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成特聘副研究员岗位要求的科研工作任务，主持国家级科研项目，参与科研奖励申报，产出高水平科研成果，积极进行成果转化；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成特聘副研究员岗位要求的教学工作任务，承担本科生课堂教学任务、指导本科生毕业设计等；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">积极参与学科平台建设，积极参加各种学术活动和国际交流；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成校院两级公共服务及学院交办的其他工作。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成岗位聘任合同中规定的其他工作。</span></p></li></ol><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">招聘条件:</span></strong></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">具有良好的思想道德素质和职业道德，品行端正，爱岗敬业，责任心强，具有较强的团结协作精神；身心健康，</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">年龄不超过35岁</span></code><span leaf="">；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">从事网络空间安全领域相关专业或方向，具有国内外高水平大学博士学位；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">具有网络空间安全学科较高学术水平，学术思想活跃，创新能力强，已在本学科领域取得一定的成就，发展潜力大，原则上应达到申报我校副研究员专业技术职务成果要求。</span></p></li></ol><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">岗位3：博士后</span></strong></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">聘用方式：博士后</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">专业：网络空间安全及相关专业和方向</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">应聘条件：博士学位</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">招聘</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">4人</span></code><span leaf=""> | 学校经费，学校经费和单位/课题组经费聘用，单位经费/课题组经费聘用</span></p></li></ul><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">岗位职责:</span></strong></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成岗位要求的科研工作任务（主持和参与科研项目、发表高水平研究成果等），积极参与学院学科建设工作，协同合作导师和研究团队开展科研工作；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">承担学校及学院规定的教学辅助及公共服务工作（指导本科生毕业设计、指导大创竞赛、实习实训带队、监考等）；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">积极参加国内外学术交流活动；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">完成校院安排的其他工作。</span></p></li></ol><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">招聘条件:</span></strong></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">具有良好的思想道德素质和职业道德，品行端正，爱岗敬业，责任心强，具有较强的团结协作精神；身心健康，</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">年龄不超过35岁</span></code><span leaf="">；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">从事网络空间安全领域相关专业或方向，具有国内外高水平大学博士学位（</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">2026年12月31日前获博士学位</span></code><span leaf="">）；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">具有网络空间安全学科较高学术水平，发展潜力大。学术思想活跃，创新能力强。</span></p></li></ol><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">二、招聘程序</span></h3><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">应聘人员通过四川大学人才招聘网站（</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf=""><a href="https://recruitment.scu.edu.cn" target="_blank">https://recruitment.scu.edu.cn</a></span></code><span leaf="">）在线应聘，填写个人应聘申请；注册时请提供准确的个人信息，因个人信息不实造成的后果由应聘者本人承担；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">学校人事处进行资格初审；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">单位招聘工作领导小组对初审通过的应聘人员进行甄选，确定参加面试人员名单；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">单位招聘工作小组、相关部处人员对候选人进行面试，党政联席会择优确定拟聘人员；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">拟聘人员进行体检、公示，无异议后上报学校；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">拟聘人员报学校审批，审批通过后转移档案、签订聘用合同、办理报到手续。</span></p></li></ol><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">三、“双百人才工程”</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">四川大学第十六批“双百人才工程”正式启动！本轮&#34;双百人才工程&#34;针对拟从校外全职引进的海内外优秀人才(截至</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">2026年3月1日未报到入职</span></code><span leaf="">的人员)。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">A计划:</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">理工医科年龄原则上不超过42周岁，人文社科原则上不超过52周岁，具有扎实学识，学术文化造诣高深，创新能力强，在开展重大原始创新和理论创新、解决关键核心技术难题和重大社会现实问题等方面取得高质量成果，作出突出贡献，获得国内外同行公认，有较大的学术影响力。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">B计划:</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">理工医科年龄原则上不超过35周岁(</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">海外引进可放宽至38周岁</span></code><span leaf="">)，人文社科原则上不超过40周岁，创新发展潜力大，在开展原始创新和理论创新、解决关键核心技术难题和重大社会现实问题等方面取得较高质量成果，在国内外同行中具有一定学术影响力。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">C计划(海外青年专项):</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">年龄</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">原则上不超过35周岁</span></code><span leaf="">，创新发展潜力大，在开展原始创新和理论创新、解决关键核心技术难题和重大社会现实问题等方面取得一定成果。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf=""><a class="normal_text_link" target="_blank" style="" href="https://mp.weixin.qq.com/s?__biz=MzI4NDAwMjYzMg==&amp;mid=2247500078&amp;idx=1&amp;sn=c06fafe18be83d24777a4cc69996bef3&amp;scene=21#wechat_redirect" textvalue="四川大学第十六批“双百人才工程”正式启动啦！" data-itemshowtype="0" linktype="text" data-linktype="2">四川大学第十六批“双百人才工程”正式启动啦！</a></span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">截至日期:</span></strong><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">2026年4月24日</span></code></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 20px;text-align: justify;"><span leaf="">四、联系方式</span></h3><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">欢迎已毕业或计划今年回蓉发展的校友与同仁，如有关于学院引才的任何疑问，请随时与我们联系。其中，</span><code style="font-size: 14px;overflow-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0px 2px;color: rgb(30, 107, 184);background-color: rgba(27, 31, 35, 0.05);font-family: &#34;Operator Mono&#34;, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">“双百人才工程”是助力个人发展与入职川大的绝佳平台</span></code><span leaf="">，诚邀您把握此次机遇。 学院官网：<a href="https://ccs.scu.edu.cn" target="_blank">https://ccs.scu.edu.cn</a></span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><span leaf="">学院人事邮箱（刘老师）: </span><code style="font-size: 14px;word-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0 2px;color: #1e6bb8;background-color: rgba(27,31,35,.05);font-family: Operator Mono, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">ccsrs<a class="wx_topic_link" topic-id="mnnwx3kv-jkbtmy" style="color: #576B95 !important;" data-topic="1" data-recommend="">#scu</a>.edu.cn (<a class="wx_topic_link" topic-id="mnnwx3kv-cwcb5w" style="color: #576B95 !important;" data-topic="1" data-recommend="">#替换为</a>@)</span></code></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><span leaf="">学院青年教师（黄老师）：</span><code style="font-size: 14px;word-wrap: break-word;padding: 2px 4px;border-radius: 4px;margin: 0 2px;color: #1e6bb8;background-color: rgba(27,31,35,.05);font-family: Operator Mono, Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">codesec<a class="wx_topic_link" topic-id="mnnwx3kv-hat5kc" style="color: #576B95 !important;" data-topic="1" data-recommend="">#scu</a>.edu.cn (<a class="wx_topic_link" topic-id="mnnwx3kv-1hurlt" style="color: #576B95 !important;" data-topic="1" data-recommend="">#替换为</a>@)</span></code></p></li></ul></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="max-width: 100%;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;white-space: normal;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;box-sizing: border-box !important;overflow-wrap: break-word !important;margin-bottom: 0px;"><blockquote style="border-color: rgb(64, 64, 64);color: rgba(0, 0, 0, 0.5);max-width: 100%;box-sizing: border-box !important;overflow-wrap: break-word !important;"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="max-width: 100%;text-align: center;box-sizing: border-box !important;overflow-wrap: break-word !important;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);-webkit-tap-highlight-color: rgba(0, 0, 0, 0);cursor: pointer;max-width: 100%;box-sizing: border-box !important;overflow-wrap: break-word !important;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a>, 有兴趣加入学术圈的请联系secdr#qq.com</span></p></blockquote></p><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>



<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=ed5eb333&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495216%26idx%3D1%26sn%3Dbcd7ba526b02d23900f886b19f806b6a">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Tue, 07 Apr 2026 09:03:00 +0800</pubDate>
    </item>
    <item>
      <title>2026年开源鸿蒙技术课题</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495207&amp;idx=1&amp;sn=3ebcd63487d8015fc8e7628e271c9dab</link>
      <description>截至26年3月，在产学研各界专家学者的深度参与下，社区成功完成了关键难题的识别与分解。</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-06 08:50</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=65650dcc&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnzwDGwFQ2HwOpjRM27My54pgSK9MxibAWQlXd3438ia8lVjCFNT8GCicGbYN9VSYhxSHYsnT3Ba8mqSm9gVY4iaGPoby6bOHsvENFo%2F0%3Fwx_fmt%3Djpeg"/></p>
  
  <p style="margin-bottom: 0px;letter-spacing: 0.578px;text-align: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.4255555555555556" data-s="300,640" data-type="png" data-w="900" style="height: 246px;width: 578px;" data-cropselx1="0" data-cropselx2="578" data-cropsely1="0" data-cropsely2="321" data-imgfileid="100009490" src="https://wechat2rss.xlab.app/img-proxy/?k=e1d3dd2d&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F6Dibw6L070WHqFxBGaQCzyTLqUunVD0DHiaxngHHyv1M90jeVsuXReYSw10gicjKE9c2qN0y3gtsicYgkeKeAW2aGg%2F640%3Fwx_fmt%3Djpeg%26from%3Dappmsg"/></span></p><div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">截至26年3月，在产学研各界专家学者的深度参与下，社区成功完成了关键难题的识别与分解。经过社区各技术领域专家与TSC评审，2026年首批开源鸿蒙技术课题现正式发布至社区。本次发布课题共计12道。详见下表：</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011546" data-ratio="1.037962962962963" style="display: block;margin: 0 auto;max-width: 100%;" data-type="png" data-w="1080" src="https://wechat2rss.xlab.app/img-proxy/?k=b656957d&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnz5FVYUmlcaUpQAM7lct2VBa1BScnZvSQqtc6tuibKwedxyyDFx6Hia3Rf08kb7NYq9ntcg44iaYTnkdv22YtLJOZcOrSQ7GibGgWQ%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">• 课题申报方式：</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">进入OH-TSC小程序→点击“课题”→点击“参与技术课题”→点击“我要申报”</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">• 课题发布时间：</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">后续申报课题通过评审后，在6月、9月、12月分批次发布到社区。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">• 以上课题具体材料获取方式：</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">进入OH-TSC小程序→点击“课题”→点击“下载课题资料”→点击“待揭榜课题”</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">• 揭榜课题方式：</span></strong></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">进入OH-TSC小程序→点击“课题”→点击“参与技术课题”→点击“我要揭榜”</span></p><h3 style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 18px;" data-pm-slice="0 0 []"><span leaf="">课题揭榜</span></h3><ol style="box-sizing: border-box;margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: rgb(0, 0, 0);list-style-type: decimal;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#34;PingFang SC&#34;, Cambria, Cochin, Georgia, Times, &#34;Times New Roman&#34;, serif;font-size: 16px;font-style: normal;font-variant-ligatures: normal;font-variant-caps: normal;font-weight: 400;letter-spacing: normal;orphans: 2;text-align: left;text-indent: 0px;text-transform: none;widows: 2;word-spacing: 0px;-webkit-text-stroke-width: 0px;white-space: normal;background-color: rgb(255, 255, 255);text-decoration-thickness: initial;text-decoration-style: initial;text-decoration-color: initial;" class="list-paddingleft-1"><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">课题揭榜者：高校OpenHarmony技术俱乐部的老师、学生，课题揭榜者即为课题负责人。</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">课题负责人：课题负责人应组建课题组，主导课题研究利社区代码合入节奏，保证研究技术原创性。</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">课题组：由一名课题负责人与其他课题组成员共同组成。</span></p></li><li style="box-sizing: border-box;"><p style="box-sizing: border-box;margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf="">揭榜时间：每年设置集中揭榜时间段。</span></p></li></ol><h4 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 18px;"><span leaf="">课题详细信息</span></h4><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.5703703703703704" data-type="png" data-w="1080" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011549" src="https://wechat2rss.xlab.app/img-proxy/?k=52963c3c&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnzibvAfgyibrVlWzxQ1SKXKC4JnicqHTJJiaTEHnXwYSqyialmeNEguV5jrGU8HEAoNbjF9icnJzoWL8asLEpgMFX6u54MQZ5xicKdfrs%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.5481481481481482" data-type="png" data-w="1080" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011547" src="https://wechat2rss.xlab.app/img-proxy/?k=ee096998&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnwjwica7Rz3B1vjJBedDaPiafBPkqDauu5bVdmMnzJVhxWQnJDSXNLyZ8QeGoeUL0DCJgXjWGibtIUWQXvWQx3MmC2g24qYuNCNk4%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.5611111111111111" data-type="png" data-w="1080" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011545" src="https://wechat2rss.xlab.app/img-proxy/?k=ed29230b&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnxNotQ3zC7gJps2HvPiboAuIgUDX0pmCF6M5gtCVaJ9uKrPvOia2Ft8kFGLvicY5lia52fL8GDuO66jOKQJgODcprVFrK1K8lFrZibs%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.5601851851851852" data-type="png" data-w="1080" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011548" src="https://wechat2rss.xlab.app/img-proxy/?k=75d9c62b&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnyOnbUYc4YmJCiawvYBgLQov5AqSRkC0iacX57FkIxT7hyd7mbRUq2PhN6ZUGniaWCNslzbHBwXnPqWgLpXCr26pRjaLqEnriba6z4%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.5564814814814815" data-type="png" data-w="1080" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011550" src="https://wechat2rss.xlab.app/img-proxy/?k=23dbc11c&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnzXel0wKmJFt6tCaBHAKedUGyBIiaticuzOvu1oPmAK4bOB1GC1w92x9PEgpgVsMv2lWfPJ2xxSN6QKxtlA9azFdPqrxzSdWJSVk%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure></div><p style="margin-bottom: 0px;letter-spacing: 0.578px;"><span leaf=""><span textstyle="" style="font-size: 16px;">原文：</span><a class="normal_text_link" target="_blank" style="" href="https://mp.weixin.qq.com/s?__biz=MzI1Mjg2NjkwOQ==&amp;mid=2247503466&amp;idx=1&amp;sn=1e35cbdb25dd4f1f11dba84af3dbe9cf&amp;scene=21#wechat_redirect" textvalue="资讯速递 | 2026年度技术课题发布" data-itemshowtype="0" linktype="text" data-linktype="2"><span textstyle="" style="font-size: 16px;">资讯速递 | 2026年度技术课题发布</span></a></span></p><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="white-space: normal; max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__43" style="margin-bottom: 0px;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);color: rgba(0, 0, 0, 0.5);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a>,</span><span leaf=""><br/></span><strong><span leaf="">有兴趣加入学术圈的请联系secdr#qq.com</span></strong></p></blockquote></p><hr style="letter-spacing: 0.578px;border-style: solid;border-right-width: 0px;border-bottom-width: 0px;border-left-width: 0px;border-color: rgba(0, 0, 0, 0.1);transform-origin: 0px 0px;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://mp.weixin.qq.com/s/UHhwHAWZTjd6sSwmDfzmKQ#rd">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=ba2814b8&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495207%26idx%3D1%26sn%3D3ebcd63487d8015fc8e7628e271c9dab">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Mon, 06 Apr 2026 08:50:00 +0800</pubDate>
    </item>
    <item>
      <title>2026年度CCF-蚂蚁科研基金（网络空间安全领域）</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495192&amp;idx=1&amp;sn=a42da0cbcc07b762f6a33e4d7465ee15</link>
      <description>基金申报截止时间为：2026年5月6日24:00（北京时间）</description>
      <content:encoded><![CDATA[<p><span>小编</span> <span>2026-04-05 16:20</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=b1e076b6&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnzer2nJeWkGEwn1julib0IEzlhMXNaoia1ao1gKiabRttpfQKEayOyPiceZlZbtRL3xHGKMlX4HWJFRd1ibPiceZMhxJ1dYjYZ6nriaYs%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>基金申报截止时间为：2026年5月6日24:00（北京时间）</p>
  <p style="margin-bottom: 0px;letter-spacing: 0.578px;text-align: justify;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.4255555555555556" data-s="300,640" data-type="png" data-w="900" style="height: 246px;width: 578px;" data-cropselx1="0" data-cropselx2="578" data-cropsely1="0" data-cropsely2="321" data-imgfileid="100009490" src="https://wechat2rss.xlab.app/img-proxy/?k=e1d3dd2d&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F6Dibw6L070WHqFxBGaQCzyTLqUunVD0DHiaxngHHyv1M90jeVsuXReYSw10gicjKE9c2qN0y3gtsicYgkeKeAW2aGg%2F640%3Fwx_fmt%3Djpeg%26from%3Dappmsg"/></span></p><div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 10px;padding-right: 10px;background-attachment: scroll;background-clip: border-box;background-color: rgba(0, 0, 0, 0);background-image: none;background-origin: padding-box;background-position-x: left;background-position-y: top;background-repeat: no-repeat;background-size: auto;width: auto;font-family: Optima, &#39;Microsoft YaHei&#39;, PingFangSC-regular, serif;font-size: 16px;color: rgb(0, 0, 0);line-height: 1.5em;word-spacing: 0em;letter-spacing: 0em;word-break: break-word;overflow-wrap: break-word;text-align: left;" data-pm-slice="0 0 []"><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: justify;text-indent: 0em;margin: 0px;padding: 8px 0px;"><span leaf="">蚂蚁集团于2020年与CCF联合发起了</span><code style="color: rgb(30, 107, 184);font-size: 14px;line-height: 1.8em;letter-spacing: 0em;background: none left top / auto no-repeat scroll padding-box border-box rgba(27, 31, 35, 0.05);width: auto;height: auto;margin: 0px 2px;padding: 2px 4px;border-style: none;border-width: 3px;border-color: rgb(0, 0, 0) rgba(0, 0, 0, 0.4) rgba(0, 0, 0, 0.4);border-radius: 4px;overflow-wrap: break-word;font-family: Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">“CCF-蚂蚁科研基金”</span></code><span leaf="">，基金连续运作6年来，致力发挥产学研的能量解决更多社会问题，受到全球学者的广泛关注和认可，累计支持额度超过6000万，吸引近1500位专家学者进行申报，支持</span><code style="color: rgb(30, 107, 184);font-size: 14px;line-height: 1.8em;letter-spacing: 0em;background: none left top / auto no-repeat scroll padding-box border-box rgba(27, 31, 35, 0.05);width: auto;height: auto;margin: 0px 2px;padding: 2px 4px;border-style: none;border-width: 3px;border-color: rgb(0, 0, 0) rgba(0, 0, 0, 0.4) rgba(0, 0, 0, 0.4);border-radius: 4px;overflow-wrap: break-word;font-family: Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">超220个科学研究项目</span></code><span leaf="">。蚂蚁基金不仅支持学者前沿探索，也积极鼓励各项科研成果的实际应用。目前，多项创新成果已成功落地，广泛应用于行业技术升级与社会问题的解决。</span></p><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: justify;text-indent: 0em;margin: 0px;padding: 8px 0px;"><span leaf="">2026年3月31日，蚂蚁集团联合CCF正式发布2026年度CCF-蚂蚁科研基金，本期基金紧密结合蚂蚁真实业务场景与人工智能发展前沿，聚焦</span><code style="color: rgb(30, 107, 184);font-size: 14px;line-height: 1.8em;letter-spacing: 0em;background: none left top / auto no-repeat scroll padding-box border-box rgba(27, 31, 35, 0.05);width: auto;height: auto;margin: 0px 2px;padding: 2px 4px;border-style: none;border-width: 3px;border-color: rgb(0, 0, 0) rgba(0, 0, 0, 0.4) rgba(0, 0, 0, 0.4);border-radius: 4px;overflow-wrap: break-word;font-family: Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">人工智能、超算与智算、网络空间安全与隐私保护、软硬协同</span></code><span leaf="">四大核心赛道，重点涵盖AI智能体、大语言模型与知识图谱、多模态融合与交互、空间计算、大模型推理加速、Token优化、大模型安全等技术方向的共44项核心议题。项目</span><code style="color: rgb(30, 107, 184);font-size: 14px;line-height: 1.8em;letter-spacing: 0em;background: none left top / auto no-repeat scroll padding-box border-box rgba(27, 31, 35, 0.05);width: auto;height: auto;margin: 0px 2px;padding: 2px 4px;border-style: none;border-width: 3px;border-color: rgb(0, 0, 0) rgba(0, 0, 0, 0.4) rgba(0, 0, 0, 0.4);border-radius: 4px;overflow-wrap: break-word;font-family: Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">合作周期为一年</span></code><span leaf="">。具体课题如下：</span></p><h4 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: block;"><span style="font-size: 18px;color: rgb(0, 0, 0);line-height: 1.5em;letter-spacing: 0em;text-align: left;font-weight: bold;display: block;"><span leaf="">方向三：网络空间安全与隐私保护</span></span></h4><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: justify;text-indent: 0em;margin: 0px;padding: 8px 0px;"><span leaf="">本基金面向人工智能安全与可信计算领域的前沿需求共发布7项前沿课题，旨在构建涵盖</span><code style="color: rgb(30, 107, 184);font-size: 14px;line-height: 1.8em;letter-spacing: 0em;background: none left top / auto no-repeat scroll padding-box border-box rgba(27, 31, 35, 0.05);width: auto;height: auto;margin: 0px 2px;padding: 2px 4px;border-style: none;border-width: 3px;border-color: rgb(0, 0, 0) rgba(0, 0, 0, 0.4) rgba(0, 0, 0, 0.4);border-radius: 4px;overflow-wrap: break-word;font-family: Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">模型安全、软件系统安全、数据安全</span></code><span leaf="">的一体化智能安全技术体系。在大模型安全层面，针对多模态触发与动态植入的后门攻击，提出鲁棒性防御方案，同时探索防蒸馏对抗机制，保障模型知识产权安全；在软件与系统安全层面，研究智能化多语言程序分析关键技术，提升跨语言代码审计的自动化水平与覆盖能力，探索基于大模型的自动化威胁验证技术；在数据安全层面，开展零知识证明友好的分组密码算法设计，为隐私计算提供高效的底层密码原语支撑；基于智能体驱动的方法，探索链上数据资产安全分析与风险治理方法的新范式；研究基于大语言模型对隐私数据语义的深度理解能力，动态构建并维护一个面向多种复杂系统的隐私数据检测与分类体系。</span></p><ol style="list-style-type: decimal;margin-top: 8px;margin-bottom: 8px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 25px;padding-right: 0px;color: rgb(0, 0, 0);" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">基于大模型的自动化威胁验证技术研究</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">面向多模态触发与动态植入的大模型后门防御技术研究</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">大模型防蒸馏对抗与核心知识资产动态防护技术研究</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">智能化多语言程序分析关键技术研究</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">零知识证明友好的分组密码设计</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">面向区块链资产安全的智能体驱动型安全分析与风险治理研究</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">面向复杂技术生态的自适应隐私数据检测系统研究</span></p></li></ol><h4 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: block;"><span style="font-size: 18px;color: rgb(0, 0, 0);line-height: 1.5em;letter-spacing: 0em;text-align: left;font-weight: bold;display: block;"><span leaf="">方向四：软硬协同</span></span></h4><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: justify;text-indent: 0em;margin: 0px;padding: 8px 0px;"><span leaf="">本基金方向是自2023年蚂蚁发布第一期“软硬协同专项”以来的连续第四期课题发布。本期基金旨在深耕“同态加密实用化”这一核心命题，精准发布5项关键技术研究课题，力求通过系统性地完善</span><code style="color: rgb(30, 107, 184);font-size: 14px;line-height: 1.8em;letter-spacing: 0em;background: none left top / auto no-repeat scroll padding-box border-box rgba(27, 31, 35, 0.05);width: auto;height: auto;margin: 0px 2px;padding: 2px 4px;border-style: none;border-width: 3px;border-color: rgb(0, 0, 0) rgba(0, 0, 0, 0.4) rgba(0, 0, 0, 0.4);border-radius: 4px;overflow-wrap: break-word;font-family: Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">同态加密技术链路</span></code><span leaf="">，在破除同态加密性能壁垒的同时，构建从算法原型到工程实践的闭环体系，从而织密隐私计算的技术版图。希望通过对软硬协同范式的持续联合研究投入，开启同态加密技术大规模产业化应用的新征程。</span></p><ol style="list-style-type: decimal;margin-top: 8px;margin-bottom: 8px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 25px;padding-right: 0px;color: rgb(0, 0, 0);" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">基于GPU的多密钥CKKS加速</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">基于多GPU的FHE加速</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">面向同态加密的CSD功耗-性能协同优化方法</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">面向全同态密文大模型推理的超节点异构近数据计算系统</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">端云协同的同态加密、转密和解密计算性能优化</span></p></li></ol><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: block;"><span style="font-size: 20px;color: rgb(0, 0, 0);line-height: 1.5em;letter-spacing: 0em;text-align: left;font-weight: bold;display: block;"><span leaf="">申报条件</span></span></h3><ol style="list-style-type: decimal;margin-top: 8px;margin-bottom: 8px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 25px;padding-right: 0px;color: rgb(0, 0, 0);" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: justify;font-weight: normal;"><span leaf="">申请者须为全球高校/科研院所在职全职教师或研究人员（博士后除外）；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: justify;font-weight: normal;"><span leaf="">在同等条件下，申请者为CCF会员为优先选择；</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: justify;font-weight: normal;"><span leaf="">能独立开展研究工作，并带领学生团队共同参与课题研究与实践。</span></p></li></ol><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: justify;text-indent: 0em;margin: 0px;padding: 8px 0px;"><span leaf="">本基金将面向符合如上条件的海内外高校及科研院所学者展开。</span></p><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: block;"><span style="font-size: 20px;color: rgb(0, 0, 0);line-height: 1.5em;letter-spacing: 0em;text-align: left;font-weight: bold;display: block;"><span leaf="">申报方式</span></span></h3><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;text-indent: 0em;margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 8px;padding-bottom: 8px;padding-left: 0px;padding-right: 0px;"><span leaf="">基金申报截止时间为：</span><code style="color: rgb(30, 107, 184);font-size: 14px;line-height: 1.8em;letter-spacing: 0em;background-attachment: scroll;background-clip: border-box;background-color: rgba(27, 31, 35, 0.05);background-image: none;background-origin: padding-box;background-position-x: left;background-position-y: top;background-repeat: no-repeat;background-size: auto;width: auto;height: auto;margin-top: 0px;margin-bottom: 0px;margin-left: 2px;margin-right: 2px;padding-top: 2px;padding-bottom: 2px;padding-left: 4px;padding-right: 4px;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgb(0, 0, 0);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 4px;border-top-right-radius: 4px;border-bottom-right-radius: 4px;border-bottom-left-radius: 4px;overflow-wrap: break-word;font-family: Consolas, Monaco, Menlo, monospace;word-break: break-all;"><span leaf="">2026年5月6日24:00（北京时间）</span></code><span leaf="">。</span></p><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;text-indent: 0em;margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 8px;padding-bottom: 8px;padding-left: 0px;padding-right: 0px;"><span leaf="">下载附件，查看、填写申报书，申请者请通过蚂蚁集团技术研究院官网(<a href="https://www.antresearch.com/Cooperation/Funds/FundsDetails/20260331001)申报。" target="_blank">https://www.antresearch.com/Cooperation/Funds/FundsDetails/20260331001)申报。</a></span></p><ul style="list-style-type: disc;margin-top: 8px;margin-bottom: 8px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 25px;padding-right: 0px;color: rgb(0, 0, 0);" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">申请人在申报前需确认所在高校/科研院所可以作为项目依托单位签署科研合作协议，申请人本人可以作为项目负责人签署项目保密协议等相关承诺文件。</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">任何针对项目申报的问题，请联系蚂蚁集团学术合作部，电子邮箱：AntResearch@antgroup.com。</span></p></li></ul><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: block;"><span style="font-size: 20px;color: rgb(0, 0, 0);line-height: 1.5em;letter-spacing: 0em;text-align: left;font-weight: bold;display: block;"><span leaf="">时间计划</span></span></h3><ul style="list-style-type: disc;margin-top: 8px;margin-bottom: 8px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 25px;padding-right: 0px;color: rgb(0, 0, 0);" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">2026年3月31日：基金发布，申报开始</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">2026年4月14/16日：蚂蚁基金课题直播宣讲</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">2026年5月6日：proposal申请截止，启动评审</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">2026年6月16日：评审结束，公布项目结果</span></p></li><li><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf="">2026年7月中旬：合同签署完毕（签署合同关注CCF-蚂蚁集团联合项目组通知）</span></p></li></ul><h3 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: block;"><span style="font-size: 20px;color: rgb(0, 0, 0);line-height: 1.5em;letter-spacing: 0em;text-align: left;font-weight: bold;display: block;"><span leaf="">详细指南内容</span></span></h3><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;text-indent: 0em;margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 8px;padding-bottom: 8px;padding-left: 0px;padding-right: 0px;"><strong style="color: rgb(0, 0, 0);font-weight: bold;background-attachment: scroll;background-clip: border-box;background-color: rgba(0, 0, 0, 0);background-image: none;background-origin: padding-box;background-position-x: left;background-position-y: top;background-repeat: no-repeat;background-size: auto;width: auto;height: auto;margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;"><span leaf="">方向三：网络空间安全与隐私保护 （支持金额不超过30万元/项）</span></strong></p><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.9564814814814815" data-type="png" data-w="1080" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011528" src="https://wechat2rss.xlab.app/img-proxy/?k=f4b29ace&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnyOHCcMRngUGR8D86bUHCLVTpBXyXjE6qVHlNtCNehxqa9C9C06qyFtvgx8dg0EcWEvIVHm704AZ3XyiaOdoXYsCop7UdL6tun8%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="1.1851851851851851" data-type="png" data-w="1080" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011530" src="https://wechat2rss.xlab.app/img-proxy/?k=9891619c&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnwIvxOhGicic9icFEknzF0nWvdSPWIVIPdsehb52W8fyolicu93ibkT1bfCpWMBmECCFFIyjK519ceTibs68hhwdft0ribw7iaISDQ3mow%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="1.1796296296296296" data-type="png" data-w="1080" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011531" src="https://wechat2rss.xlab.app/img-proxy/?k=cf4c9ed7&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnydCY44E62R0icpKrsojWu5niabsY4iattBswJ3liciae9zy9OuOBzom1KbrY8iawh540Ty5dnyWeBvR4pXTXonI5JpnjgPccB32MHp8%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="1.225" data-type="png" data-w="1080" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011532" src="https://wechat2rss.xlab.app/img-proxy/?k=60fec77e&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnw1vJHEudvueNibV5AcTew256TaBcUicWVkYcq45VO9BMeNbUmOia8K4IF7DBD3I1ec31ZgQ6esVBz9Oz4AoF8clqJ5ybyHN08Y7k%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.8564814814814815" data-type="png" data-w="1080" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011529" src="https://wechat2rss.xlab.app/img-proxy/?k=a512541f&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnzPll3hPWx6edibWjyibo7B4qH19YicibDM26uvpfGlE2n81AODscrgeNmdiaRCMSAWibHYJOtruWZtmEnf4xOEKt6F1kiaxQqSAIYaPA%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;text-indent: 0em;margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 8px;padding-bottom: 8px;padding-left: 0px;padding-right: 0px;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="1.5784543325526932" data-type="png" data-w="854" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;" data-imgfileid="100011537" src="https://wechat2rss.xlab.app/img-proxy/?k=9ae5afe6&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnwTicZOiab6S4cicaI1VmBG7X5zh60kpvKMOuujSl5QpXxdOj9zRz4ngViclqrfC8RXgGzGY58fY6x4IAk0VTJmVXnewPmhkXofLd0%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.5602836879432624" data-type="png" data-w="846" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;" data-imgfileid="100011533" src="https://wechat2rss.xlab.app/img-proxy/?k=a6e20dc6&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnzmV5r8YkBDmoz6oXfwDljtZPu8gQnkKNRrfm8w7YrFHeuOV8GuSjiabL91yuhwNOBrga5vuLdicicFhU2OSgQ2Gt7ibAzPiayvzWHg%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="1.2890995260663507" data-type="png" data-w="844" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011536" src="https://wechat2rss.xlab.app/img-proxy/?k=c574d76f&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrny2Fj2iaXlFicOoic95Ab6KnNyGore1ujiaoLtxTRaY8RwLnMW8s9H95V2qFDyoUkAFAgyEJA5GibxH7ibhb7U8CoiaTtib8hOnkrXQJsA%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;text-indent: 0em;margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 8px;padding-bottom: 8px;padding-left: 0px;padding-right: 0px;"><strong style="color: rgb(0, 0, 0);font-weight: bold;background-attachment: scroll;background-clip: border-box;background-color: rgba(0, 0, 0, 0);background-image: none;background-origin: padding-box;background-position-x: left;background-position-y: top;background-repeat: no-repeat;background-size: auto;width: auto;height: auto;margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;"><span leaf="">方向四：软硬协同（支持金额不超过40万元/项）</span></strong></p><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.9225352112676056" data-type="png" data-w="852" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011534" src="https://wechat2rss.xlab.app/img-proxy/?k=73ccafb6&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnyynNHJwonfRTm2TGdcqHlfKgFloiagiaPLiaTFsGtiaTbnwS1MHvKt9icaYIGm27YD61oMgeIwOvticoQzUgGl3HnR5x00Ag8Nqor3I%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="1.0094117647058825" data-type="png" data-w="850" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011535" src="https://wechat2rss.xlab.app/img-proxy/?k=ff1ea667&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnxWdsaI1DaEYRiaZFVcK1DDicItcJsFElIAwnJV3H8UERwic0JY12QZdTt1Tu4tI3HWZIhod0VvPgq0N8icvr9oUJLe3tcOWWYjqgo%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.9635535307517085" data-type="png" data-w="878" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011539" src="https://wechat2rss.xlab.app/img-proxy/?k=18fa5a29&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnw9YMKr8mwYW2QKt8EWUdzblt4oTkibwB6ibxQOvNcLt9BSdcupia2ia9YYOsjXKdhibC6ibqwHbmhjWaSianaatHicCNyWoBFWXLjegQs%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="1.1967213114754098" data-type="png" data-w="854" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011541" src="https://wechat2rss.xlab.app/img-proxy/?k=e98564ad&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnwDZsRPw6YwxMuMfhqTlm2npicHyJ7VsDehptkGiaIJtvbgoRDTLn8eVgicUjGxApXDsEVpxFV6GUFMCQYfF7Q2pDgUWSq6gcttBE%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><figure data-tool="mdnice编辑器" style="margin-top: 10px;margin-bottom: 10px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 0px;padding-right: 0px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="1.2214452214452214" data-type="png" data-w="858" style="display: block;margin-top: 0px;margin-right: auto;margin-bottom: 0px;margin-left: auto;max-width: 100%;border-top-style: none;border-bottom-style: none;border-left-style: none;border-right-style: none;border-top-width: 3px;border-bottom-width: 3px;border-left-width: 3px;border-right-width: 3px;border-top-color: rgba(0, 0, 0, 0.4);border-bottom-color: rgba(0, 0, 0, 0.4);border-left-color: rgba(0, 0, 0, 0.4);border-right-color: rgba(0, 0, 0, 0.4);border-top-left-radius: 0px;border-top-right-radius: 0px;border-bottom-right-radius: 0px;border-bottom-left-radius: 0px;object-fit: fill;box-shadow: rgba(0, 0, 0, 0) 0px 0px 0px 0px;" data-imgfileid="100011540" src="https://wechat2rss.xlab.app/img-proxy/?k=c8d8be01&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnzdDfnHpdjZEs7Fic2EOg0JXMdKkfsthSyt7FqWDKmQaCQb8Fyxrq0X2UbLQ7JeUlly57TPliatmpDq8f2QVEcdqEARbHXaiaxRGk%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p data-tool="mdnice编辑器" style="color: rgb(0, 0, 0);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;text-indent: 0em;margin-top: 0px;margin-bottom: 0px;margin-left: 0px;margin-right: 0px;padding-top: 8px;padding-bottom: 8px;padding-left: 0px;padding-right: 0px;"><span leaf="">官方链接：</span></p><ul style="list-style-type: disc;margin-top: 8px;margin-bottom: 8px;margin-left: 0px;margin-right: 0px;padding-top: 0px;padding-bottom: 0px;padding-left: 25px;padding-right: 0px;color: rgb(0, 0, 0);" class="list-paddingleft-1"><li style="font-style:italic;font-size:14px;"><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf=""><span textstyle="" style="font-size: 14px;font-style: italic;text-decoration: underline;"><a href="https://www.ccf.org.cn/Collaboration/Enterprise_Fund/News/my/2026-03-31/870286.shtml" target="_blank">https://www.ccf.org.cn/Collaboration/Enterprise_Fund/News/my/2026-03-31/870286.shtml</a></span></span></p></li><li style="font-style:italic;font-size:14px;"><p style="margin-top: 5px;margin-bottom: 5px;color: rgb(1, 1, 1);font-size: 16px;line-height: 1.8em;letter-spacing: 0em;text-align: left;font-weight: normal;"><span leaf=""><span textstyle="" style="font-size: 14px;font-style: italic;text-decoration: underline;"><a href="https://www.antresearch.com/Cooperation/Funds/FundsDetails/20260331001" target="_blank">https://www.antresearch.com/Cooperation/Funds/FundsDetails/20260331001</a></span></span></p></li></ul></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="white-space: normal; max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__43" style="margin-bottom: 0px;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);color: rgba(0, 0, 0, 0.5);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a>,</span><span leaf=""><br/></span><strong><span leaf="">有兴趣加入学术圈的请联系secdr#qq.com</span></strong></p></blockquote></p><hr style="letter-spacing: 0.578px;border-style: solid;border-right-width: 0px;border-bottom-width: 0px;border-left-width: 0px;border-color: rgba(0, 0, 0, 0.1);transform-origin: 0px 0px;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://www.ccf.org.cn/Collaboration/Enterprise_Fund/News/my/2026-03-31/870286.shtml">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=4b28665a&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495192%26idx%3D1%26sn%3Da42da0cbcc07b762f6a33e4d7465ee15">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Sun, 05 Apr 2026 16:20:00 +0800</pubDate>
    </item>
    <item>
      <title>康考迪亚大学 | OCR-APT：基于子图异常检测与大语言模型的审计日志APT攻击事件重构</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495175&amp;idx=1&amp;sn=bb079ef94385e793ef1adfcde60c5539</link>
      <description>本文提出OCR-APT，试图实现从系统审计日志到APT攻击故事重建的端到端分析</description>
      <content:encoded><![CDATA[<p>原创 <span>龙函城</span> <span>2026-04-02 18:30</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=b184ea8f&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_jpg%2F8dQ2T4EPrnwhFvGXhd1CkeIBYic4SYgPHy9s3l4fkZmJXniaZMNxRMgibib4GdxgbwRke5zZiaLQUUAj0ZVMKX6PaBbVffXBcV9266PjGmBrNqVg%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>本文提出OCR-APT，试图实现从系统审计日志到APT攻击故事重建的端到端分析</p>
  <div data-tool="mdnice编辑器" data-website="https://www.mdnice.com" style="font-size: 16px;color: black;padding: 0 10px;line-height: 1.6;word-spacing: 0px;letter-spacing: 0px;word-break: break-word;word-wrap: break-word;text-align: left;font-family: Optima-Regular, Optima, PingFangSC-light, PingFangTC-light, &#39;PingFang SC&#39;, Cambria, Cochin, Georgia, Times, &#39;Times New Roman&#39;, serif;" data-pm-slice="0 0 []"><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011522" data-ratio="0.22592592592592592" style="display: block;margin: 0 auto;max-width: 100%;" data-type="png" data-w="1080" src="https://wechat2rss.xlab.app/img-proxy/?k=f849c84c&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnxamrX4iaUzzVNmhNjJjL5TZ5EAT50u7O6YYKnc5SjaVzuibp9M2mjctYNgzib3iaHt7Vian7SaqJ4tvK8WibEsVTp8P3kxPJLEAs2nE%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><blockquote style="border: none;display: block;font-size: 0.9em;overflow: auto;overflow-scrolling: touch;border-left: 3px solid rgba(0, 0, 0, 0.4);background: rgba(0, 0, 0, 0.05);color: #6a737d;padding-top: 10px;padding-bottom: 10px;padding-left: 20px;padding-right: 10px;margin-bottom: 20px;margin-top: 20px;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;color: black;line-height: 26px;text-align: justify;"><em style="font-style: italic;color: black;"><span leaf="">原文标题：OCR-APT: Reconstructing APT Stories from Audit Logs using Subgraph Anomaly Detection and LLMs</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">原文作者：Ahmed Aly, Essam Mansour, Amr Youssef</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">发表会议：CCS ’25</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">笔记作者：龙函城</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">主编：黄诚@安全学术圈</span></em></p></blockquote><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">研究概述</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">高级持续性威胁（Advanced Persistent Threat, APT）具有长期潜伏和多阶段推进等特点，传统基于单点告警或线性日志的分析方式往往难以完整恢复其攻击过程。在这一背景下，溯源图（Provenance Graph）由于能够将进程、文件、网络流等实体及其交互关系组织为具有因果语义的图结构，逐渐成为APT检测与调查的重要技术路线。相比普通日志表示，溯源图更适合揭示攻击行为之间的依赖关系与传播路径。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，现有基于溯源图的APT分析方法仍存在两个突出问题。一方面，许多异常检测方法只能输出节点级、时间窗级或整图级告警，容易出现粒度过粗、上下文不足和误报率较高的问题；另一方面，已有攻击调查系统往往依赖预定义的兴趣点、已知攻击实体或规则线索来启动分析，且输出多为复杂子图、事件序列或低层日志片段，缺乏连贯的攻击叙事能力，分析人员仍需手工梳理攻击从何开始、如何推进以及涉及哪些关键IOC。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">针对上述问题，本文提出OCR-APT，试图实现从系统审计日志到APT攻击故事重建的端到端分析。该方法主要由两部分组成：第一部分是基于图神经网络的子图异常检测模块，用于从大规模溯源图中识别与APT相关的异常节点和异常子图；第二部分是基于大语言模型的攻击调查模块，用于将这些异常子图进一步转化为包含APT阶段、关键IOC及行为上下文的人类可读攻击报告。在检测层面，作者提出OCRGCN，将关系图卷积网络（RGCN）与one-class SVM结合，通过结构关系和行为模式建模异常，而不是依赖文件路径、IP地址等脆弱属性；在调查层面，作者设计了分阶段、带验证机制的RAG流程，以减少幻觉并提高报告可信度。</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.21316033364226136" data-type="png" data-w="1079" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011523" src="https://wechat2rss.xlab.app/img-proxy/?k=5cca7607&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnwrgj3kibREBC21pJicHxpkWvQVpsKuQ3As6rhE3OnKkTWDwc1TtibcniaPdiaBhcut21EJBcicvTggheDpF5AqUQ3rfAiaW573YKPJibs%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: center;"><span leaf="">图 1 OCR-APT 系统整体架构</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如图1所示，OCR-APT的整体流程从审计日志出发，首先完成溯源图构建、行为特征提取与图编码；随后通过OCRGCN识别异常节点，并围绕这些节点构造异常子图；最后再将异常子图输入LLM-based attack investigator，生成带有APT阶段、关键IOC和上下文解释的人类可读攻击报告。该框架体现了本文的核心思想，即将APT分析划分为子图异常检测和攻击故事重建两个阶段，从而实现从低层系统遥测到高层攻击叙事的映射。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">贡献分析</span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点1</span></strong><span leaf=""> ：提出面向APT检测的子图级异常检测模型OCRGCN</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文提出了OCRGCN，将RGCN与one-class SVM相结合，用于在溯源图中识别异常节点及APT相关异常子图。OCRGCN 强调基于结构关系和行为模式进行建模，从而降低伪相关问题对检测结果的影响，提升对未知APT行为的鲁棒性。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点2</span></strong><span leaf=""> ：提出基于LLM的APT攻击调查与故事重建方法</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文进一步引入了一个LLM-driven attack investigator，用于从检测得到的异常子图中重建攻击过程，并生成简洁、可读的攻击报告。该部分的重点是将异常子图组织为带有APT阶段、关键IOC和行为上下文的攻击叙事，使系统输出更贴近安全分析人员的实际调查需求。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点3</span></strong><span leaf=""> ：设计了分阶段且带验证机制的RAG调查流程</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文并没有直接让大语言模型一步完成完整攻击报告生成，而是将攻击调查过程拆分为IOC提取、IOC验证、子图摘要、综合报告生成和上下文增强等多个子任务，并在中间环节加入自动验证机制，过滤掉并不存在于原始子图中的IOC，从而减少幻觉问题并提升最终报告的可信度。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点4</span></strong><span leaf=""> ：构建了检测与调查一体化的完整APT分析系统 论文最终将子图异常检测与LLM攻击调查整合为一个完整系统OCR-APT，使其能够从系统审计日志出发，依次完成溯源图构建、异常节点识别、异常子图构造、告警排序以及攻击报告生成。</span></p></div></li></ul><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">代码分析</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">代码链接</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf=""><a href="https://github.com/CoDS-GCS/OCR-APT" target="_blank">https://github.com/CoDS-GCS/OCR-APT</a></span></p></li></ul><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">使用类库分析</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">该项目整体采用Python与Bash作为主要实现方式，系统一方面通过GraphDB和SPARQL完成审计日志到溯源图的存储与查询，另一方面结合PyTorch Geometric风格的图表示学习流程完成异常节点与异常子图检测；同时，还引入LLM与向量检索相关组件，用于生成攻击调查报告。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码实现难度与工作量评估</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从实现角度来看，该项目难度属于中等偏上。它的难点不在单一算法本身，而在于需要把多个环节稳定串联起来，包括原始审计日志转换、RDF溯源图构建、图编码、one-class GNN训练、异常节点识别、异常子图构造以及 LLM 调查报告生成。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码关键实现功能</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（1）审计日志到溯源图转换模块。该模块主要负责将原始CSV格式的审计日志转换为RDF表示，并导入GraphDB，为后续图查询和图学习建模提供统一的数据基础。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（2）图编码与特征提取模块。该模块用于将溯源图进一步编码为PyG可处理的数据结构，并提取后续GNN模型所需的节点与关系信息，为异常检测做准备。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（3）基于GNN表示学习与one-class SVM的异常检测模块。该模块对应OCRGCN的核心实现，主要先利用RGCN学习溯源图中节点的结构与行为表示，再结合one-class SVM学习正常节点分布，并在测试阶段识别异常节点。它是整个系统完成APT异常检测的核心部分。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（4）异常子图构造模块。该模块根据检测出的异常节点继续构造异常子图，并筛选出更具有调查价值的可疑子图，从而避免只输出孤立异常点。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（5）LLM攻击调查与报告生成模块。该模块利用大语言模型对异常子图进行进一步分析，并生成简洁、可读的攻击调查报告，是论文中攻击故事重建能力的关键实现部分。</span></p></div></li></ol><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文点评</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">总体来看，这篇论文是一项较有价值的APT检测研究。它的亮点不只是提升异常检测能力，更在于进一步关注APT分析中的实际需求，即不仅要发现异常，还要尽可能讲清楚异常背后的攻击过程。为此，作者将子图异常检测与LLM攻击调查结合起来，使系统能够从审计日志出发，进一步生成带有APT阶段、关键IOC和行为上下文的攻击报告，这一点比仅输出异常节点或异常子图更贴近真实分析场景。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从方法设计上看，论文整体思路比较清晰。前半部分利用溯源图和OCRGCN识别异常节点与异常子图，后半部分再通过LLM完成攻击故事重建，这种先检测、后调查的设计较符合APT分析流程。尤其是作者没有直接让大模型一步生成完整结论，而是采用分阶段、带验证机制的RAG流程来减少幻觉，这使方法在可解释性和可信度上更进一步。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，这篇论文也存在一定局限。例如，系统整体依赖审计日志和溯源图质量，若底层数据不完整，检测和重建效果都会受到影响；同时，方法涉及图数据库、图学习和LLM生成多个环节，整体实现链路较长，部署与复现成本相对较高。总体而言，本文的主要价值在于它展示了APT检测系统如何从“发现异常”进一步走向“解释异常”，对APT检测与攻击调查结合方向具有较强参考意义。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文文献</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">Büchel, Marvin, et al. &#34;{SoK}: Automated {TTP} Extraction from {CTI} Reports–Are We There Yet?.&#34; 34th USENIX security symposium (USENIX Security 25). 2025.</span></p></div><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://arxiv.org/pdf/2510.15188">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=db3883c7&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495175%26idx%3D1%26sn%3Dbb079ef94385e793ef1adfcde60c5539">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Thu, 02 Apr 2026 18:30:00 +0800</pubDate>
    </item>
    <item>
      <title>奥尔登堡大学 | SoK：从 CTI 报告中自动化抽取 TTP——我们真的做到这一步了吗？</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495169&amp;idx=1&amp;sn=2ffc09f2445ae8a081c07f1955e76be0</link>
      <description>本文不仅帮助读者更清楚地理解自动化TTP抽取领域的主要方法路线，也提醒研究者：这一方向未来的突破，可能不仅依赖更强的模型能力，也依赖更高质量的数据和更合理的任务定义。</description>
      <content:encoded><![CDATA[<p>原创 <span>龙函城</span> <span>2026-03-26 18:31</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=9e84a968&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnw7eALOueibYiaAOTlT2szibq2G1WYfek0fE3CZkY8M4n4KD39jfd5dafyOqWdcaGJcD6mibwDrWu0br2F8EIznsickiaX4xS1icAvE04%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>本文不仅帮助读者更清楚地理解自动化TTP抽取领域的主要方法路线，也提醒研究者：这一方向未来的突破，可能不仅依赖更强的模型能力，也依赖更高质量的数据和更合理的任务定义。</p>
  <p style="margin-bottom: 0px;letter-spacing: 0.578px;text-wrap: wrap;text-align: left;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.2833333333333333" data-type="png" data-w="1080" style="display: block;margin: 0 auto;max-width: 100%;" data-imgfileid="100011516" src="https://wechat2rss.xlab.app/img-proxy/?k=c3a75a6a&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnz6atzqOhtJD1F6X1IS1CiaUYd6knUDRr9vkGMAc2J4XVxM8qVMpgvxaky94XFNmjFoiaSg2yVl8b9nMOChTReO2FpqSuIqH1oyc%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><blockquote style="border: none;display: block;font-size: 0.9em;overflow: auto;overflow-scrolling: touch;border-left: 3px solid rgba(0, 0, 0, 0.4);background: rgba(0, 0, 0, 0.05);color: #6a737d;padding-top: 10px;padding-bottom: 10px;padding-left: 20px;padding-right: 10px;margin-bottom: 20px;margin-top: 20px;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;color: black;line-height: 26px;text-align: justify;"><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 15px;">原文标题：SoK: Automated TTP Extraction from CTI Reports – Are We There Yet?</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 15px;">原文作者：Marvin Büchel, Tommaso Paladini, Stefano Longari, Michele Carminati, and others.</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 15px;">发表会议：34th USENIX Security Symposium</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 15px;">笔记作者：龙函城</span></span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf=""><span textstyle="" style="font-size: 15px;">主编：黄诚@安全学术圈</span></span></em></p></blockquote><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">研究概述</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">随着网络威胁情报（Cyber Threat Intelligence, CTI）在威胁发现、攻击溯源和安全运营中的作用不断增强，安全分析工作已经从早期面向IoC的简单匹配，逐步扩展到对攻击者行为模式的深入理解。其中，战术、技术与过程（Tactics, Techniques, and Procedures, TTPs）能够更系统地描述攻击者的攻击目标、实施方式及具体操作过程，因此成为CTI分析中的关键内容。当前，MITRE ATT&amp;CK已成为描述和组织TTP的事实标准框架，但大量CTI报告仍以非结构化自然语言形式存在，导致从报告中自动抽取并映射TTP仍然是一项复杂、耗时且高度依赖人工经验的任务。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">近年来，围绕CTI报告自动化TTP抽取，研究界已经提出了大量基于自然语言处理的方法。总体来看，现有方法大致可以归纳为三类：NER方法、分类方法和生成式LLM方法。其中，NER 方法主要识别文本中显式出现的攻击技术表达，具有一定可解释性，但对隐式语义的处理能力有限；分类方法通常将TTP抽取建模为多标签分类问题，能够从句子整体语义出发判断对应的ATT&amp;CK技术标签，但较为依赖高质量标注数据；生成式LLM方法则借助大语言模型的理解与生成能力，直接根据CTI文本生成相关技术编号或名称，形式更灵活，但也容易受到提示设计、输出不稳定和幻觉问题的影响。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，虽然相关研究数量不断增加，现有成果之间却往往难以进行公平比较。这是因为不同工作通常采用不同的数据集、不同的TTP本体、不同的任务设定和评测指标，很多方法还依赖私有或不可公开的CTI语料，从而使得研究社区很难准确判断：当前自动化TTP抽取到底发展到了什么水平，哪些技术路线真正有效，距离实际可用还有多远。针对这一问题，本文对该领域进行系统梳理与实验比较。作者首先系统归纳已有自动化TTP抽取工作，然后在统一设定下复现并比较不同技术路线，从而更公平地分析各类方法的优势、局限及适用场景。</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img data-aistatus="1" class="rich_pages wxw-img" data-ratio="0.6449612403100775" data-type="png" data-w="645" style="display:block;margin:0 auto;max-width:100%;width:497px;height:321px;" data-imgfileid="100011517" src="https://wechat2rss.xlab.app/img-proxy/?k=c485f6b5&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnw5KYbstFiahUiazlaCuJwcS6WdxSmJwViadDkx4yibyPrRay0kjnl7HWODicicExXhtDUeaVQACYHPzAoIZe1e4euBjicMRqDlHtTTPY%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: center;"><span leaf=""><span textstyle="" style="font-size: 14px;">图 1 自动化 TTP 抽取方法在开放场景下的性能比较 </span></span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如图1所示，作者进一步在更贴近真实应用的开放场景下比较了不同方法在多种ATT&amp;CK技术规模下的表现。实验结果表明，随着候选技术标签数量不断增多，不同方法的性能都会出现明显变化，且生成式方法并没有在真实设定下稳定优于传统分类方法。作者据此指出，当前自动化TTP抽取研究似乎已经遇到某种阶段性性能瓶颈：一方面，现有方法在受限条件下可以取得较好结果；但另一方面，一旦进入更开放、更复杂、更加接近实际应用的场景，其整体可靠性仍然不足，尚难支撑高精度、全覆盖的自动化ATT&amp;CK标注任务。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">基于系统化文献分析与统一实验比较，论文最终得出几个关键结论：第一，自动化TTP抽取目前还无法真正可靠地覆盖整个ATT&amp;CK框架；第二，在更真实的设定下，传统NLP方法和较小的分类模型并未被现代生成式方法显著超越，某些情况下反而更加稳定；第三，制约该领域进一步发展的关键因素，可能并不只是模型结构本身，而更在于高质量标注数据的缺乏、类别分布不平衡以及ATT&amp;CK技术标签之间天然存在的语义歧义和边界模糊。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">贡献分析</span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点1</span></strong><span leaf=""> ：系统梳理了CTI报告自动化TTP抽取领域的研究现状</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文首先对现有的TTP自动化抽取研究进行了系统化整理，从文献层面对已有工作进行了归纳与分类，将相关研究概括为NER、Classification和Generative LLM三类主流技术路线，并进一步分析了各类方法的适用条件、优势与局限。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点2</span></strong><span leaf=""> ：实现了多类代表性方法的统一技术复现与比较框架</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">为了解决现有研究难以横向比较的问题，论文进一步对不同技术路线进行了统一实现。作者在一致的ATT&amp;CK本体、统一的数据集和相同评测标准下，对NER、分类方法以及生成式方法进行了系统比较，相比单纯罗列已有工作，这种统一技术实现使论文具备了更强的实证分析价值。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点2</span></strong><span leaf=""> ：揭示了自动化TTP抽取领域当前面临的核心瓶颈</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文通过统一实验表明，现有自动化TTP抽取方法在更真实的开放场景下仍存在明显局限，生成式方法也未表现出对传统方法的稳定优势。作者进一步指出，限制该领域持续突破的关键因素，不仅在于模型本身的能力边界，更在于高质量标注数据的缺乏、类别分布不平衡以及ATT&amp;CK技术标签之间普遍存在的语义歧义与边界模糊问题。</span></p></div></li></ul><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">代码分析</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf=""><span textstyle="" style="font-size: 12px;">代码链接：<a href="https://doi.org/10.5281/zenodo.15608555." target="_blank">https://doi.org/10.5281/zenodo.15608555.</a></span></span></p><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">使用类库分析</span></strong><span leaf="">：该项目整体属于典型的“多路线方法复现+统一评测”型研究代码框架。仓库主要分为 NER、classification、generation 三个子项目，分别对应论文中的三类主流方法，同时还包含数据集、文献爬取和对比实验等辅助模块。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码实现难度与工作量评估</span></strong><span leaf="">：从实现角度来看，该项目难度属于 中等偏上。它的难点不在于单一模型本身，而在于需要同时复现多条技术路线，并在统一的数据集、ATT&amp;CK 本体和评测标准下进行比较。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码关键实现功能</span></strong><span leaf="">：结合代码结构与复现实验说明，其核心功能可概括为以下4个模块：</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（1）NER方法实现模块。该模块对应论文中的传统NLP路线，主要围绕文本预处理、术语归一化、规则匹配和ATT&amp;CK技术映射来完成TTP抽取。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（2）分类模型实现模块。该模块主要实现基于文本表示的TTP分类，包括有监督多标签分类和基于语义相似度的分类思路。其核心功能应包括文本编码、标签预测、多标签输出和开放场景下的技术匹配。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（3）生成式LLM实验模块。该模块对应论文中的生成式方法路线，主要围绕大语言模型完成TTP抽取实验。除了基础提示推理外，这部分还应支持few-shot、RAG和监督微调等实验方式，并将模型输出进一步映射到ATT&amp;CK技术名称。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（4）数据集模块。从整体仓库结构来看，作者专门提供了统一的数据集组织部分，用于管理TRAM2、AnnoCTR、增强数据以及生成式模型训练所需的指令数据。</span></p></div></li></ol><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文点评</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">此论文并没有停留在对已有工作的简单罗列上，而是通过统一实验，对CTI报告自动化TTP抽取领域的不同技术路线进行了较为系统的比较。作者发现，在封闭标签空间下，分类方法和生成式方法能够取得较好的结果；但在更接近真实应用的开放场景中，不同方法的性能都会明显下降，而且基于大语言模型的生成式方法并没有表现出对其他方法的稳定优势。相反，以NER为代表的传统NLP方法在某些更现实的设定下仍具有一定竞争力。这一结论对当前相关研究具有较强的参考价值。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">这篇论文没有把问题简单归结为模型还不够强，而是进一步指出该领域当前真正的瓶颈更可能在于高质量标注数据不足、类别分布不平衡，以及ATT&amp;CK技术标签之间存在天然的语义歧义和边界模糊。也就是说，自动化TTP抽取之所以难以持续突破，并不只是模型结构的问题，还与任务定义方式、标签体系设计以及数据覆盖范围密切相关。这使得论文不只是做了方法比较，也对该方向未来的研究重点提出了较清晰的判断。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，本文也存在一定局限。例如，作者所依赖的公开数据集本身仍不够理想，部分数据集覆盖的技术范围有限，部分数据集又存在一定标签噪声；同时，论文更强调不同技术路线的统一比较，因此未必完整纳入各原始方法中的特定优化策略。总体而言，本文不仅帮助读者更清楚地理解自动化TTP抽取领域的主要方法路线，也提醒研究者：这一方向未来的突破，可能不仅依赖更强的模型能力，也依赖更高质量的数据和更合理的任务定义。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文文献</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">Büchel, Marvin, et al. &#34;{SoK}: Automated {TTP} Extraction from {CTI} Reports–Are We There Yet?.&#34; 34th USENIX security symposium (USENIX Security 25). 2025.</span></p><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://www.usenix.org/system/files/usenixsecurity25-buechel.pdf">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=372346b3&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495169%26idx%3D1%26sn%3D2ffc09f2445ae8a081c07f1955e76be0">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Thu, 26 Mar 2026 18:31:00 +0800</pubDate>
    </item>
    <item>
      <title>中国科学院信息工程研究所 | 基于大语言模型的API参数安全规则生成与API误用检测</title>
      <link>https://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247495163&amp;idx=1&amp;sn=4cfc8079192ebcc8de2fd692a0d2c5b8</link>
      <description>本文提出了一个结构完整、思路清晰的方法框架，展示了大语言模型在 API 安全规则生成与误用检测中的应用潜力。</description>
      <content:encoded><![CDATA[<p>原创 <span>龙函城</span> <span>2026-03-19 18:30</span> <span style="display: inline-block;">四川</span></p>






  
  <p><img src="https://wechat2rss.xlab.app/img-proxy/?k=a2260e80&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_jpg%2F8dQ2T4EPrnwxyv085a9puADWVWXicEhSb2nBOhgZNicq2rvx9b0tbmanfUoftKJ3dQmqQhPIMdCDKib57IiaYKicia000z9micvE1BuO3uxNKcp0XA%2F0%3Fwx_fmt%3Djpeg"/></p>
  <p>本文提出了一个结构完整、思路清晰的方法框架，展示了大语言模型在 API 安全规则生成与误用检测中的应用潜力。</p>
  <p style="margin-bottom: 0px;letter-spacing: 0.578px;text-wrap: wrap;text-align: left;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011510" data-ratio="0.31203703703703706" style="display: block;margin: 0 auto;max-width: 100%;" data-type="png" data-w="1080" src="https://wechat2rss.xlab.app/img-proxy/?k=02627a7b&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fmmbiz_png%2F8dQ2T4EPrnz2TPHVl7HadU4G9e8wgVtvL2tmZibqubs5YWiaAhUbJ6wB0w4rkQkmrndeia2D3Tia1cicX6XDF5VGWJflQYYqZwfFkFyEgJRcBkek%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></p><blockquote style="border: none;display: block;font-size: 0.9em;overflow: auto;overflow-scrolling: touch;border-left: 3px solid rgba(0, 0, 0, 0.4);background: rgba(0, 0, 0, 0.05);color: #6a737d;padding-top: 10px;padding-bottom: 10px;padding-left: 20px;padding-right: 10px;margin-bottom: 20px;margin-top: 20px;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;color: black;line-height: 26px;text-align: justify;"><em style="font-style: italic;color: black;"><span leaf="">原文标题：Generating API Parameter Security Rules with LLM for API Misuse Detection</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">原文作者：Jinghua Liu , Yi Yang, Kai Chen, Miaoqian Lin</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">发表会议：NDSS Symposium 2025</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">笔记作者：龙函城</span></em><span leaf=""><br/></span><em style="font-style: italic;color: black;"><span leaf="">主编：黄诚@安全学术圈</span></em></p></blockquote><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">研究概述</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">随着现代软件开发对第三方库 API 的广泛依赖，开发者在调用库函数时不仅要保证功能实现正确，还需要遵循相应的安全使用规范，尤其是参数层面的安全约束。论文将研究重点放在 API Parameter Security Rules（APSRs） 上。APSRs 是 API 安全规则中的一种常见类型，主要用于描述参数取值、参数状态以及相关操作应满足的安全条件。若违反这些规则，就可能引发空指针解引用、内存破坏、拒绝服务等安全问题。作者指出，在已有 API 误用案例中，约71%与 APSRs 违规直接相关，这说明 APSRs 是 API 安全分析中的关键内容。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">然而，APSR 的构建并不容易。现有方法通常依赖 API 文档、调用代码模式或传统静态分析来提取规则，但这些方法普遍存在文档不完整、调用样本有限、分析规则依赖人工设计等问题，因此容易遗漏重要约束。相比之下，API 源码更接近真实实现过程，而大语言模型又具备较强的代码理解和文本生成能力，因此论文尝试利用 LLM 直接从 API 源码中自动生成 APSRs，以进一步支持 API 误用检测。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，作者进一步发现，直接使用 LLM 生成 APSRs 效果并不好。一方面，LLM 可能会因为幻觉生成错误规则；另一方面，模型也容易输出过于宽泛、模糊的规则。论文的实验表明，若直接依赖 LLM 生成 APSRs，其精度只有11.9%，说明这一问题十分突出。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">针对上述问题，本文提出 GPTAid 框架，用于自动生成准确且具体的 API 参数安全规则，并进一步将这些规则应用到 API 误用检测中。GPTAid 的核心思想不是直接相信 LLM 的输出，而是构建一个“生成—验证—细化—检测”的闭环流程，其框架如图1所示。</span></p><figure data-tool="mdnice编辑器" style="margin: 0;margin-top: 10px;margin-bottom: 10px;display: flex;flex-direction: column;justify-content: center;align-items: center;"><span leaf=""><img class="rich_pages wxw-img" data-aistatus="1" data-imgfileid="100011511" data-ratio="0.8328947368421052" style="display:block;margin:0 auto;max-width:100%;width:457px;height:381px;" data-type="png" data-w="760" src="https://wechat2rss.xlab.app/img-proxy/?k=ace46860&amp;u=https%3A%2F%2Fmmbiz.qpic.cn%2Fsz_mmbiz_png%2F8dQ2T4EPrnxIwVcFNr9PR5oQr2G3sbzfhicjibJyib6UHtQS319tf1NEc7ETGfjjVw64jPMjUWo0iaxufGAk5qUzvetAWpYHQz1tQqR5bJb3vPU%2F640%3Fwx_fmt%3Dpng%26from%3Dappmsg"/></span></figure><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: center;"><em style="font-style: italic;color: black;"><span leaf="">图 1 GPTAid 框架整体流程</span></em></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">如图1所示，GPTAid 整体包括四个阶段：首先基于 API 源码生成原始 APSRs；随后围绕每条规则生成能够正确调用 API 的 Right Code 和故意违反规则的 Violation Code，并通过运行时错误反馈验证规则是否正确；接着，系统利用代码差分分析，将原始规则进一步细化为更具体、更清晰的 APSRs；最后，再把这些规则转换为检测逻辑，用于在真实应用中发现 API 误用问题。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">该方法的关键在于两点。第一，论文提出了执行反馈校验机制，即通过执行违规代码并观察运行时错误，过滤掉那些看似合理但实际上不正确的规则。第二，论文提出了代码差分细化机制，通过比较正确代码和违规代码之间的关键差异，把原本模糊的自然语言规则细化成更明确的安全约束，从而便于后续自动生成检测代码。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">实验结果表明，GPTAid 在200个 API 上生成 APSRs 的精度达到92.3%，召回率达到71.0%；与已有方法相比，它能够生成更多类型的规则，并在47个真实应用中发现210个未知 API 误用问题，其中150个已得到开发者确认。总体来看，这篇论文展示了将大语言模型与程序分析技术结合，用于 API 安全规则生成与误用检测的有效性和实用价值。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">贡献分析</span></h2><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点1</span></strong><span leaf="">：提出了基于 LLM 的 APSRs 自动生成框架 GPTAid</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文提出 GPTAid，用于从 API 源码中自动生成 API 参数安全规则。与直接使用大语言模型不同，GPTAid 重点面向两个核心难点：一是 LLM 容易生成错误规则，二是容易生成过于宽泛、缺乏可执行性的规则。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点2</span></strong><span leaf="">：设计了执行反馈校验与代码差分细化两项关键技术</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">为了提升生成规则的可靠性，论文提出了 execution feedback-checking 机制，通过生成违规调用代码并结合运行时错误反馈来验证 APSRs 的正确性；同时提出 code differential analysis，利用正确代码与违规代码之间的关键差异，将原本模糊、笼统的规则进一步细化为更具体、更清晰的安全约束。这两项技术共同构成了 GPTAid 的核心方法基础。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点3</span></strong><span leaf="">：证明了生成规则在真实 API 误用检测中的有效性</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文不仅完成了 APSRs 的自动生成，还将这些规则进一步用于 API 误用检测，并证明了其实用价值。实验表明，GPTAid 生成的规则能够支撑后续检测流程，并且检测效果优于现有先进方法。此外，基于这些生成规则，作者共发现210个未知安全 bug，其中150个已被开发者确认，说明该方法不仅在规则生成层面有效，在真实漏洞发现任务中也具有较强应用价值。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">贡献点4</span></strong><span leaf="">：发现了大量文档缺失规则，并总结了实践性建议</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">论文在8个常用库的实验中发现，在生成的579条 APSRs 中，有61.3%在现有 API 文档中没有对应描述，这说明 GPTAid 在一定程度上能够补充和丰富 API 文档中的安全约束信息。与此同时，论文还进一步总结了两类实践建议：一类是面向 API 开发者的文档改进建议，另一类是面向 LLM 使用过程的 prompt 设计建议，从而增强了研究的工程指导意义。</span></p></div></li></ul><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">代码分析</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><span leaf="">代码链接</span></p><ul style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: disc;" class="list-paddingleft-1"><li style="font-size:16px;"><p style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: justify;color: rgb(1, 1, 1);font-weight: 500;"><span leaf=""><span textstyle="" style="font-size: 16px;"><a href="https://github.com/icy17/GPTAid/" target="_blank">https://github.com/icy17/GPTAid/</a></span></span></p></li></ul><ol style="margin-top: 8px;margin-bottom: 8px;padding-left: 25px;color: black;list-style-type: decimal;" class="list-paddingleft-1"><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0;line-height: 26px;color: black;"><strong style="font-weight: bold;color: black;"><span leaf="">使用类库分析</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">该项目整体采用 Python 作为主要实现语言，属于典型的“LLM 调用 + 程序分析 + 动态验证 + 规则细化”型研究代码框架。系统一方面调用大语言模型完成 API 参数安全规则生成、正确调用代码生成、违规调用代码生成以及规则细化；另一方面结合源码解析工具、运行时检测工具和静态检测工具，对生成结果进行验证与落地应用。整体来看，这个项目并不涉及复杂的模型训练，而是更侧重于将论文中的方法流程工程化实现。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码实现难度与工作量评估</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从实现角度来看，该项目难度属于中等偏上。它的难点不在于单一算法本身，而在于需要把论文中的多个环节稳定串联起来，包括目标库源码预处理、API 安全规则生成、正确与违规调用样例构造、运行时错误验证以及最终规则细化与检测。</span></p></div></li><li><div style="margin-top: 5px;margin-bottom: 5px;line-height: 26px;text-align: left;color: rgb(1,1,1);font-weight: 500;"><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><strong style="font-weight: bold;color: black;"><span leaf="">代码关键实现功能</span></strong></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（1）源码预处理与信息抽取模块。该部分主要负责对目标库源码进行分析，提取 API 函数定义、参数信息以及相关调用关系，为后续大语言模型生成规则提供必要上下文。它相当于整个系统的数据准备阶段，为后面的规则生成和验证打基础。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（2）API 参数安全规则生成模块。该模块围绕目标 API 及其参数构造提示信息，引导大语言模型生成初始的 API 参数安全规则。其作用是让模型从源码层面理解参数约束，而不是仅依赖文档或调用样本，从而为后续检测提供候选规则。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（3）正确调用代码与违规调用代码生成模块。为了验证规则是否合理，系统会围绕目标 API 自动生成能够正常调用的代码样例，以及违反特定规则的调用样例。该模块是把自然语言形式的规则落到可执行程序层面，使得后续能够通过真实运行结果来检验规则质量。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（4）执行反馈验证模块。该部分是整个系统的重要环节。系统通过执行生成出的违规代码，并结合运行时错误检测机制来判断某条规则是否真正有效。相比仅依赖模型文本输出，这种方式能够更可靠地过滤掉错误规则或无意义规则，是提升 APSRs 精度的关键步骤。</span></p><p style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">（5）规则细化与误用检测模块。在得到初步验证后的规则后，系统还会进一步对规则进行细化，使其从较为模糊的表述变成更具体、可解释、可检测的安全约束。最终，这些规则会被转化为检测逻辑，用于在真实应用程序中发现 API 误用问题。</span></p></div></li></ol><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文点评</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">总体来看，这篇论文是一项较有价值的 LLM 与程序分析结合的安全研究工作。它的重点并不只是利用大语言模型生成 API 参数安全规则，而是进一步针对直接使用 LLM 时容易出现的错误规则和规则过于泛化两个问题，设计了执行反馈校验和代码差分细化两项机制，从而提升了规则的准确性和可用性。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">从实验结果来看，论文具有较强说服力。GPTAid 在 APSRs 生成任务上取得了 92.3% 的精度和 71.0% 的召回率，并进一步在真实应用中发现了210个未知安全问题，其中150个 已被开发者确认。这说明该方法不仅在规则生成层面有效，而且具备较强的实际漏洞发现能力。与此同时，论文还发现大量生成规则在现有 API 文档中没有对应描述，说明 GPTAid 在补充 API 安全约束信息方面也具有现实意义。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">不过，这篇论文也存在一定局限。它的验证方式较大程度上依赖运行时错误反馈，因此对于那些不会立即触发异常的 API 误用问题，检测能力可能有限。此外，系统在规则到检测逻辑的转换过程中仍带有一定人工总结和模板化处理成分，说明其自动化程度还有进一步提升空间。</span></p><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">总体而言，本文提出了一个结构完整、思路清晰的方法框架，展示了大语言模型在 API 安全规则生成与误用检测中的应用潜力。它也说明，在软件安全分析场景下，LLM 更适合与程序验证和分析技术协同使用，而不是单独作为黑盒生成工具。</span></p><h2 data-tool="mdnice编辑器" style="margin-top: 30px;margin-bottom: 15px;padding: 0px;font-weight: bold;color: black;font-size: 22px;"><span leaf="">论文文献</span></h2><p data-tool="mdnice编辑器" style="font-size: 16px;padding-top: 8px;padding-bottom: 8px;margin: 0px;line-height: 26px;color: black;text-align: justify;"><span leaf="">Liu J, Yang Y, Chen K, Lin M. Generating API Parameter Security Rules with LLM for API Misuse Detection[C]//Network and Distributed System Security Symposium (NDSS). 2025.</span></p><p data-tool="mdnice编辑器" data-website="https://www.mdnice.com" data-style="max-width: 100%; font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif; letter-spacing: 0.544px; white-space: normal; caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); font-size: 16px; text-align: left; background-color: rgb(255, 255, 255); box-sizing: border-box !important; overflow-wrap: break-word !important;" class="js_darkmode__40" style="margin-bottom: 0px;text-wrap: wrap;background-color: rgb(255, 255, 255);color: rgb(0, 0, 0);font-family: -apple-system, BlinkMacSystemFont, &#34;Helvetica Neue&#34;, &#34;PingFang SC&#34;, &#34;Hiragino Sans GB&#34;, &#34;Microsoft YaHei UI&#34;, &#34;Microsoft YaHei&#34;, Arial, sans-serif;letter-spacing: 0.544px;caret-color: rgb(0, 0, 0);font-size: 16px;text-align: left;"><blockquote style="border-color: rgb(64, 64, 64);"><p data-darkmode-bgcolor="rgb(36, 36, 36)" data-darkmode-original-bgcolor="rgb(255, 255, 255)" data-darkmode-color="rgba(230, 230, 230, 0.5)" data-darkmode-original-color="rgba(0, 0, 0, 0.5)" style="text-align: center;"><span leaf=""><a class="normal_text_link" target="_blank" style="color: rgb(125, 144, 169);cursor: pointer;" href="http://mp.weixin.qq.com/s?__biz=MzU5MTM5MTQ2MA==&amp;mid=2247484475&amp;idx=1&amp;sn=2c91c6a161d1c5bc3b424de3bccaaee0&amp;chksm=fe2efbb0c95972a67513c3340c98e20c752ca06d8575838c1af65fc2d6ddebd7f486aa75f6c3&amp;scene=21#wechat_redirect" textvalue="安全学术圈招募队友-ing" data-itemshowtype="0" linktype="text" data-linktype="2">安全学术圈招募队友-ing</a> </span><span leaf=""><br/></span><span leaf="">有兴趣加入学术圈的请联系 </span><span style="color: rgb(255, 0, 0);"><strong><span leaf="">secdr#qq.com</span></strong></span></p></blockquote></p><hr style="border-style: solid;border-width: 1px 0 0;border-color: rgba(0,0,0,0.1);-webkit-transform-origin: 0 0;-webkit-transform: scale(1, 0.5);transform-origin: 0 0;transform: scale(1, 0.5);"/><p class="mp_profile_iframe_wrp" nodeleaf=""><mp-common-profile class="js_uneditable custom_select_card mp_profile_iframe" data-pluginname="mpprofile" data-nickname="安全学术圈" data-alias="secquan" data-from="1" data-headimg="http://mmbiz.qpic.cn/mmbiz_png/6Dibw6L070WFvZRQiafv3iccicic1dIYUEQ1ZzLh1a10l7tfw7zkWkRbY9kEPBwX2NiadOrwFl9a48as9qiayp3eOgDUQ/0?wx_fmt=png" data-signature="分享安全方面的论文写作、会议发表、基金申请方面的资料。" data-id="MzU5MTM5MTQ2MA==" data-is_biz_ban="0"></mp-common-profile></p><p style="display: none;"><mp-style-type data-value="3"></mp-style-type></p>


<p><a href="https://arxiv.org/pdf/2409.09288">阅读原文</a></p>
<p><a href="https://wechat2rss.xlab.app/link-proxy/?k=8c37cd35&amp;r=1&amp;u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%3F__biz%3DMzU5MTM5MTQ2MA%3D%3D%26mid%3D2247495163%26idx%3D1%26sn%3D4cfc8079192ebcc8de2fd692a0d2c5b8">跳转微信打开</a></p>
]]></content:encoded>
      <pubDate>Thu, 19 Mar 2026 18:30:00 +0800</pubDate>
    </item>
  </channel>
</rss>