Skip to content

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

Oct 12 · Monday

All papers

Found 128 papers
Filter3

Attack type

Jailbreaks15Prompt injection112Poisoning and backdoors20Extraction and theft3Detection evasion4

Attack technique

Gradient and representation optimization3Automated search21Multi-turn escalation3Encoding and obfuscation6Defense-targeted4Dormant triggers7

Model risks

Dangerous agent actions6

Defense type

Model hardening6Inference-time intervention3Detection and filtering17Instruction–data separation26Permissions and approval25Monitoring and auditing15

Impact

Harmful content8Unauthorized actions73Data and privacy leaks15Output manipulation26Oversight failure4

System type

LLM applications3AI Agent124coding agent8web agent3Multi-agent systems26

Attacker access

Black-box7

Attack surface / entry point

User input122Web pages and interfaces69Documents and messages79Tool responses105Tool and skill descriptions62Code repositories and dependencies29Retrieval corpus35Memory writes40Inter-agent messages28Images, audio and video20Training data3Weights and gradients4Distribution channels44Inference services and hardware3

Layer

Prompt layer3Application layer128

System component

MCP and skills20Memory19Guardrails and judges13Monitors15Reasoning chains3Permissions and sandboxing29
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 247papers have no tags yet and are not included in filtered results.
Attack surface / entry point：Tool responsesAttack surface / entry point：Inter-agent messagesLayer：Application layerClear all

Another 247papers do not have category tags yet and are not included in filtered results.

  1. AttacksarXiv 2610.12292

    The paper proposes an option channel attack: changing one option label can make an agent guardrail allow a prohibited operation

    Proposing an option channel attack that renames option labels so that agent guardrails allow prohibited operations

    arXiv
    2610.12292
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Attacker
    Black-box
    Evaluated models
    LAYA-TD, LAYA-EN, LAYA-ML and others5

    AbstractThe authors' evaluation indicates that typed models should not serve as final safety arbiters; they are susceptible to off-policy text and option name manipulation, and confidence filtering cannot resist attacks.

    Paper analysisRead analysis
  4. DefensesarXiv 2610.09793

    用时序逻辑监控工具 Agent

    提出 MFOTL 离线时序验证以监控工具 Agent 的危险动作链

    arXiv
    2610.09793
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Attacker
    Black-box
    Abstract形式化验证可有效识别多步攻击，但其有效性高度依赖轨迹的可信历史记录。

    这篇论文对工具调用大语言模型智能体的安全监控进行了形式化时序验证研究。

    Paper analysisRead analysis
  5. AttacksarXiv 2610.08951

    ASPIRE：面向 LLM Agent 的开放式提示注入红队引擎

    提出 ASPIRE 行为图搜索，开放式发现 Agent 提示注入漏洞

    arXiv
    2610.08951
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Gemini-3.7-Flash, DeepSeek-V4-Flash, Gemini-2.5-flash
    AbstractASPIRE 把提示注入红队从载荷优化改成行为图引导的开放式漏洞发现。

    ASPIRE 是给搭建方用的智能体提示注入红队引擎，目标是开放式画出可利用的行为路径，而不是在固定任务上打磨一条载荷。

    Paper analysisRead analysis
  6. DefensesarXiv 2610.08668

    SBW：面向 LLM 智能体的语义行为水印方法

    提出 SBW，为 LLM Agent 轨迹做抗改写与伪造的语义行为水印

    arXiv
    2610.08668
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Qwen2.5-3B-Instruct, LLaMA-3.2-3B, Qwen2.5-7B and others5
    AbstractSBW 在语义簇上做溯源水印，改写检测高于精确符号，链式重放仍开放。

    SBW 是免训练的智能体行为水印，把所有者标识放在离散动作选择上，而不是输出 token。

    Paper analysisRead analysis
  7. 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
  10. Evaluation and benchmarksarXiv 2610.07089

    研究者提出统一滥用监控基准，覆盖分解攻击与提示注入约 6200 条轨迹

    提出危害窗口与区间指标，统一评测分解攻击和提示注入下的智能体滥用监视

    arXiv
    2610.07089
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Gemma 4 26B-A4B, Qwen3.6 27B, Qwen3.5 27B and others12

    Abstract统一智能体滥用监视框架，动作视角兼顾两类威胁但危害定位仍存瓶颈。

    Paper analysisRead analysis
  11. DefensesarXiv 2610.05640

    Research reveals that the CaMeL defense fails in multi-agent systems and proposes multi-CaMeL

    提出 multi-CaMeL，在智能体调用边界隔离可信指令与非可信数据

    arXiv
    2610.05640
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Claude Fable 5, GPT-5.5, GPT-4.1 and others6

    Abstract论文针对多智能体中单体防御不组合的边界清洗问题，提出 multi-CaMeL 协议彻底阻断了 MultiAgentDojo 上的注入攻击。

    Paper analysisRead analysis
  12. InterpretabilityarXiv 2610.05295

    研究用因果追踪区分间接提示注入中的可读信号与有效干预

    用因果追踪区分间接提示注入中的角色可读性与有效干预

    arXiv
    2610.05295
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Qwen-2.5-7B, Qwen-2.5-1.5B, Llama-3.1-8B and others8
    Abstract作者称源角色先于强工具选择效应可读，有效编辑还取决于位置、长度和方向来源。

    在权重固定的白盒模型上，源角色可以先被读出。沿相关方向的编辑是否改变行为，要在目标位点上另测。。

    Paper analysisRead analysis
  13. DefensesarXiv 2610.05163

    Researchers construct segmented prompt injection on Claude Code and Codex and release a boundary action audit benchmark

    提出 PAA 方法，在边界动作前审计长流程智能体的分阶段提示注入

    arXiv
    2610.05163
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Claude Code (v2.1.172, Claude Sonnet 5), Codex CLI (v0.144.0, GPT-5.5)
    Abstract论文针对长流程注入提出边界动作审计与 PAA，在保持低误阻的同时实现高拦截率。

    本研究针对长流程智能体在执行任务时面临的分阶段间接提示注入威胁，提出了在动作生效前进行拦截的边界动作审计框架与路径对齐归因方法。

    Paper analysisRead analysis
  17. DefensesarXiv 2610.02861

    论文提出在 Kubernetes 上保护 AI Agent 的七层纵深防御框架与参考架构

    提出 Kubernetes 运维智能体的七层纵深防御框架与参考架构

    arXiv
    2610.02861
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Abstract把运维智能体当不可信租户，用七层控制在完全攻破假设下约束行动。

    把 Kubernetes 上的运维智能体当作不可信租户，在假定提示注入已完全控制其工具选择的前提下，用基础设施重建数据面与控制面的边界。要解决的是：智能体读日志、注解、工单或 MCP 工具描述时，这些内容可以变成它带着自身凭据执行的命令。模型对齐和分类器只降低概率。

    Paper analysisRead analysis
  18. DefensesarXiv 2610.02349

    MIRROR：面向多智能体通信的法定人数完整性防御

    提出 MIRROR，用多路径摘要多数表决保障多智能体通信完整性

    arXiv
    2610.02349
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    Gemma-4-31B-it, Gemini 2.5 Pro
    AbstractMIRROR 靠多数路径。

    MIRROR 是 LLM 多智能体通信层上的多路径法定人数完整性原语，只处理不透明字节，不解释消息内容。要处理的是 Agent-in-the-Middle：中介合法终结 TLS 后可改在途消息，而语义校验有推理开销和误拦，签名又要求私钥管理。做法是生成结束后复制同一规范化载荷。

    Paper analysisRead analysis
  19. DefensesarXiv 2610.02005

    贝叶斯辩证论证（BDA）：面向持续对抗下多 LLM 委员会的可校准聚合

    提出 BDA，按辩证动作可靠性加权聚合多 LLM 议会以抵抗持续对抗

    arXiv
    2610.02005
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    AbstractBDA 用零额外调用的可靠性加权，给出可校准后验，并在持续对抗下反转不可靠席位。

    BDA 是一种不另调 LLM 的多 LLM 议会聚合方法，同时处理置信度校准和持续不可靠席位。

    Paper analysisRead analysis
  20. Evaluation and benchmarksarXiv 2610.03448

    Study retests 15 prompt injection detectors: benchmark scores poorly predict agent deployment performance

    重测提示注入检测器在 Agent 工具输出上的检出与任务阻断

    arXiv
    2610.03448
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Horizon-Labs, Wolf Defender, Prismor-1.5B and others15

    Abstract论文揭示公开基准无法反映智能体检测器表现，检测能力源于输入格式相似而非通用防御，建议基于工具输出与低误报评测。

    Paper analysisRead analysis
Load more