Skip to content

Translation in progress Some entries still include text in the original language.

Oct 12 · Monday

All papers

Found 8 papers
Filter5

Attack type

Jailbreaks11Prompt injection21Poisoning and backdoors8Malicious use5

Attack technique

Dormant triggers3

Impact

Unauthorized actions5

System type

Models and APIs6LLM applications3AI Agent21coding agent4Multi-agent systems3Embodied systems and robotics1

Attack surface / entry point

User input8Documents and messages5Tool and skill descriptions19Code repositories and dependencies5Memory writes10Inter-agent messages4Images, audio and video4Training data6Weights and gradients6Distribution channels21

Layer

Application layer8

System component

Memory4
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 233papers have no tags yet and are not included in filtered results.
Attack type：Poisoning and backdoorsSystem type：coding agentSystem type：Multi-agent systemsSystem type：Embodied systems and roboticsAttack surface / entry point：User inputClear all

Another 233papers do not have category tags yet and are not included in filtered results.

  1. DefensesarXiv 2610.07657

    DEFER1 多智能体防御研究报告受控测试中的攻击成功率下降

    实现 DEFER，以确定性检查优先拦截多智能体流水线的内容攻击

    arXiv
    2610.07657
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Qwen3-235B-A22B-Instruct-2507, gpt-oss-120b, Llama-4-Scout-17B-16E-Instruct and others5
    AbstractDEFER 把确定性检查放在 judge 之前。

    DEFER 用来测量多智能体授权流水线里，确定性规则在何处结束、LLM judge 从何处开始。

    Paper analysisRead analysis
  2. Evaluation and benchmarksarXiv 2609.39146

    MADBench：多智能体辩论安全性评测基准发布

    提出 MADBench，评测多智能体辩论是吸收还是放大对抗影响

    arXiv
    2609.39146
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    gpt-4o, GPT-5, GPT-6 Astra
    AbstractMADBench 显示辩论能吸收部分准确率攻击，同时放大未授权读写。

    MADBench 是一个评测多智能体辩论安全性的基准，用来看攻击是被辩论吸收还是被放大。

    Paper analysisRead analysis
  3. AttacksarXiv 2609.38270

    VirusCascade：劫持 LLM 推荐智能体的协同反思机制

    提出 VirusCascade，联合操纵商品描述与交互轨迹以劫持推荐智能体协同反思

    arXiv
    2609.38270
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    AgentCF, AgentSEQ, AgentRAG and others4
    Abstract作者把反思写回视为推广攻击入口，并用联合操纵和诊断探针研究其持续与传播。

    VirusCascade 是针对推荐智能体协作反思的 black-box targeted promotion 攻击研究。 攻击者只改商品描述和少量账号的普通交互，试图让目标商品获得更多良性用户曝光。

    Paper analysisRead analysis
  4. Evaluation and benchmarksarXiv 2609.32616

    Cave-Bench：虚假指控下 LLM 智能体最高 60.06% 的运行会破坏正确工作

    提出 CAVE-BENCH，评测智能体在虚假指责下破坏已完成的正确工作

    arXiv
    2609.32616
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Claude-Sonnet-5, Claude-Opus-5, GPT-5.6-Sol and others14

    Abstract论文提出了评估智能体在虚假指责下破坏已有正确成果的 CAVE-BENCH，发现强模型易推翻已知证据，提出了门控缓解方案。

    Paper analysisRead analysis
  5. DefensesarXiv 2608.21101

    ClawSentry：面向自主 LLM Agent 的渐进式多层安全监控网关

    提出 ClawSentry 多层监控网关，在技能准入和运行时拦截恶意技能与隐藏指令

    arXiv
    2608.21101
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Codex/GPT-5.4, Codex/GPT-5.5, Claude Code/GLM-5.1 and others6
    AbstractClawSentry 用包准入、三级运行时门控和反绕过，压低 skill 攻击并保住干净任务。

    ClawSentry 是挂在智能体运行时之外的安全监督网关，用来同时盖住 skill 执行的多个生命周期边界，并把已被拒绝的效果记到会话里。

    Paper analysisRead analysis
  7. AttacksarXiv 2608.09577

    ElasticBack：通过触发词与规则耦合在 LLM Agent 技能中植入条件后门

    提出 ElasticBack，在单个 Agent 技能中植入条件后门

    arXiv
    2608.09577
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    GLM-5.2, MiniMax-M3, GPT-5.4 and others4
    AbstractElasticBack 用冻结规则加遗传搜索演化 trigger，在单 skill 上做条件后门。

    ElasticBack 是放在 LLM agent 单个 skill 里的条件后门攻击。。

    Paper analysisRead analysis
You have reached the end