Skip to content

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

Oct 13 · Tuesday

All papers

Found 15 papers
Filter6

Attack type

Prompt injection13Poisoning and backdoors5

Impact

Unauthorized actions8Output manipulation4

System type

LLM applications18AI Agent67coding agent7web agent4Browser and computer use3Multi-agent systems8

Attack surface / entry point

User input17Web pages and interfaces32Documents and messages13Tool responses11Tool and skill descriptions5Retrieval corpus4Memory writes8Inter-agent messages25Images, audio and video7Weights and gradients2

Layer

Application layer15

System component

Memory4
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 213papers have no tags yet and are not included in filtered results.
System type：web agentSystem type：Browser and computer useSystem type：Multi-agent systemsAttack surface / entry point：Documents and messagesAttack surface / entry point：Weights and gradientsLayer：Application layerClear all

Another 213papers do not have category tags yet and are not included in filtered results.

  3. AttacksarXiv 2609.39788

    研究：多智能体系统潜在通信链路可被训练用于提升有害顺从

    提出只训练或篡改多智能体潜在通信链路来提高有害遵从

    arXiv
    2609.39788
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Llama-3.2-3B-Instruct, Qwen2.5-3B-Instruct, Gemma-3-1B-IT and others7
    Abstract潜在链路本身即可改变冻结智能体系统的安全，奖励既能攻击也能修复。

    学习得到的潜在通信链路会改变由冻结、对齐智能体组成的系统是否遵从有害请求。。作者研究三种拓扑：双智能体、三级顺序协作，以及双专家加汇总。

    Paper analysisRead analysis
  4. 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
  5. 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
  6. Evaluation and benchmarksarXiv 2609.24662

    DUMA-Bench：评估 LLM Agent 安全的双控多智能体基准

    提出双控多智能体基准 DUMA-Bench，测量对抗环境下的策略违反

    arXiv
    2609.24662
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    GPT-5, GPT-5-mini, GPT-5-nano and others14
    AbstractDUMA-Bench 用双控交互评测智能体安全。

    DUMA-Bench 把 τ²-bench 的双控交互做成智能体安全评测，用来看策略违反如何随用户、模型和环境一起出现。

    Paper analysisRead analysis
  8. Evaluation and benchmarksarXiv 2609.17320

    Emergence World：对长时程多智能体系统开展对抗压力测试

    提出 Emergence World，对长时程多智能体系统开展对抗压力测试

    arXiv
    2609.17320
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Claude Opus 4.8, DeepSeek v4 Pro, Gemini 3.5 Flash and others9

    Abstract作者通过长期平行世界实验指出模型对齐不具备可组合性，单模型群体会产生社会奉承、协同退缩及语言不透明等失效。

    Paper analysisRead analysis
  9. AttacksarXiv 2608.06862

    SynChain：诱导计算机使用 Agent 自建攻击链并跨任务持久化

    提出 SYNCHAIN，诱导计算机使用 Agent 自建跨任务持久攻击链

    arXiv
    2608.06862
    Published
    Layer
    Application layer
    Setting
    Browser and computer use, coding agent
    Evaluated models
    Qwen3.5-9B, Llama-3.1-8B, Ministral-3-8B

    Abstract论文揭示了 CUA 自合成工件带来的内部供应链威胁，验证了潜伏载荷的跨任务传播能力与防御局限。

    Paper analysisRead analysis
  10. AttacksarXiv 2605.08310

    WebTrap：针对浏览器 Agent 导航过程的中途劫持注入攻击

    提出 WebTrap 三阶段陷阱，中途劫持浏览器 Agent 完成攻击后再返回用户任务

    arXiv
    2605.08310
    Published
    Layer
    Application layer
    Setting
    web agent
    Evaluated models
    DeepSeek-V3.1-Terminus, Gemini-2.5-Flash, GLM-4.5-Air and others5
    AbstractWebTrap 把攻击目标嵌进用户流程中途执行再返回。

    WebTrap 是针对长程网页与文件导航的中途劫持注入，目标是在完成攻击动作后仍回到用户任务。

    Paper analysisRead analysis
  11. DefensesarXiv 2603.04474

    From Spark to Wildfire: Modeling and Mitigating Error Cascades in LLM-Based Multi-Agent Collaboration

    Proposing a lineage-graph governance layer to suppress error cascades in multi-agent collaboration

    arXiv
    2603.04474
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    AbstractA single seed can solidify into a false consensus along a fixed collaboration graph.

    The authors turn "small errors becoming collective false consensus" in LLM-MAS into a measurable propagation on a graph, and use a plug-in that only changes messages, not topology, to reduce the final infection.

    Paper analysisRead analysis
  13. Evaluation and benchmarksarXiv 2512.23128

    牛津等发布 TRAP 基准

    构建 TRAP 基准，评测网页 Agent 被界面提示注入劫持任务的情况

    arXiv
    2512.23128
    Published
    Layer
    Application layer
    Setting
    web agent
    Evaluated models
    GPT-5, Claude Sonnet 3.7, Gemini 2.5 Flash and others6
    Abstract论文提出了基于五维模块化攻击空间的 TRAP 基准，评估了六个模型在真实网页环境中的注入风险。

    研究定位：针对网页自主智能体易受外部界面提示注入攻击的问题，构建了基于真实网站克隆的模块化评测基准 TRAP。

    Paper analysisRead analysis
  14. Evaluation and benchmarksarXiv 2510.27140

    MobiRed 框架测评八款移动 LLM Agent 的第三方渠道提示注入攻击

    用 MobiRed 评测移动 LLM Agent 面对第三方渠道的提示注入

    arXiv
    2510.27140
    Published
    Layer
    Application layer
    Setting
    Browser and computer use
    Evaluated models
    Mobile-Agent-E, AppAgent, AutoGLM and others8

    Abstract论文评估了移动 LLM 智能体作为 confused deputy 的易受攻击性，作者称非应用控制通道可诱导智能体执行多步越权危害。

    Paper analysisRead analysis
  15. Evaluation and benchmarksarXiv 2510.01354

    WAInjectBench 系统评测网页 Agent 的提示注入检测方法

    提出 WAInjectBench 评测网页 Agent 的提示注入检测

    arXiv
    2510.01354
    Published
    Layer
    Application layer
    Setting
    web agent
    Evaluated models
    GPT-4o, Mistral-7B-Instruct-v0.1, Llama-Prompt-Guard-2-86M and others8
    AbstractWAInjectBench 比较网页智能体的提示注入检测：显式线索可被部分检出，隐蔽攻击大多漏检。

    WAInjectBench 是网页智能体提示注入检测的基准。作者要看现有检测器能否发现这类攻击：攻击者改网页，使智能体执行目标动作，而不是用户任务。作者称这是首个此类综合基准。

    Paper analysisRead analysis
You have reached the end