Skip to content

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

Oct 12 · Monday

All papers

Found 24 papers
Filter5

Attack type

Prompt injection21Poisoning and backdoors5

Defense type

Detection and filtering6Instruction–data separation6Permissions and approval4

Impact

Unauthorized actions11Output manipulation10

System type

Models and APIs5LLM applications10AI Agent142coding agent8web agent23Browser and computer use8Multi-agent systems6

Attack surface / entry point

User input43Web pages and interfaces16Documents and messages26Tool responses10Tool and skill descriptions5Code repositories and dependencies4Retrieval corpus27Memory writes8Inter-agent messages25Images, audio and video10Training data5Weights and gradients3Distribution channels3Inference services and hardware4

Layer

Application layer24

System component

RAG3Guardrails and judges7Permissions and sandboxing3Vision5Multimodal5
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.
System type：LLM applicationsSystem type：Browser and computer useSystem type：Multi-agent systemsAttack surface / entry point：Web pages and interfacesAttack surface / entry point：Tool responsesClear all

Another 233papers 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
  3. AttacksarXiv 2610.11932

    A Fudan team measures the citation fragility of AI search: ordinary posts can enter citations and answer text

    Measures citation fragility in AI search, demonstrating that ordinary public posts can enter citations and answers.

    arXiv
    2610.11932
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    Grok, Doubao, ChatGPT and others10

    AbstractThe authors claim the paper quantifies vulnerabilities in AI search caused by citation preferences and low posting thresholds; tests show public web posts can penetrate AI citations, and they call for auditing the citation supply chain.

    Paper analysisRead analysis
  4. 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
  5. 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
  6. DefensesarXiv 2610.03089

    COBRA: A Dual-LLM Architecture for Defending Computer-Using Agents Against Branch-Jumping Attacks

    提出 COBRA 双 LLM 架构，防御计算机使用 Agent 的分支转向攻击

    arXiv
    2610.03089
    Published
    Layer
    Application layer
    Setting
    Browser and computer use
    Evaluated models
    gpt-5, UI-TARS-1.5-7B, claude-sonnet-5 and others6

    Abstract提出分支转向防御架构 COBRA，通过事前能力约束与边界代理消除分支转向攻击并兼顾效用。

    Paper analysisRead analysis
  8. Evaluation and benchmarksarXiv 2610.06898

    DIBench：面向移动 GUI 智能体多候选选择决策完整性的安全基准

    用 DIBench 评测移动 GUI 智能体被 UI 注入后的决策完整性

    arXiv
    2610.06898
    Published
    Layer
    Application layer
    Setting
    Browser and computer use
    Evaluated models
    Qwen2.5-VL(32B), Qwen3-VL(32B), GUI-Owl(32B) and others7

    Abstract论文提出移动智能体决策完整性基准 DIBench，发现欺诈 UI 注入诱导目标选择并推高完成率，通用防御收益不稳定。

    Paper analysisRead analysis
  9. Evaluation and benchmarksarXiv 2609.33102

    ORBIT：基于 UK AISI Inspect 的多智能体安全评测框架

    提出 ORBIT 框架，评测多智能体防御对合谋等威胁的效果

    arXiv
    2609.33102
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    GPT-5.4, Claude Sonnet 4.6, Qwen3-235B and others5

    AbstractORBIT 用六层配置比较多智能体防御，逐动作监控不迁移到合谋。

    Paper analysisRead analysis
  10. Analysis and theoryarXiv 2609.23215

    研究审计 Active Inference Agent 的 LLM 解释器失败模式

    审计 Active Inference Agent 上 LLM 解释器的监督叙述失败

    arXiv
    2609.23215
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    GPT-4o, Claude-3-Opus, Gemini
    Abstract三组触发下解释流畅但错误，所提缓解都未被评估。

    作者审计的是挂在智能体上的运行时解释器，而不是智能体本身。操作员读到的是信念与动作的自然语言叙述。作者要问哪些可复现触发会让这段叙述失败，以及失败痕迹是什么样。

    Paper analysisRead analysis
  14. Evaluation and benchmarksarXiv 2609.02316

    Counter-GEO-Bench：现有护栏难挡 GEO 虚假信息

    构建 COUNTER-GEO-BENCH，评测生成式搜索中的 GEO 虚假信息防御

    arXiv
    2609.02316
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    Gemma-4-31B-IT, Qwen-3.5-35B-A3B, Llama-4-Scout-17B-16E and others8

    Abstract论文构建了针对 GEO 虚假信息的首个防御基准，揭示通用护栏的失效与反作用，并提出了有效的轻量对比检测基线。

    Paper analysisRead analysis
  15. Evaluation and benchmarksarXiv 2609.05535

    Mind2Web-Injection：面向网页 Agent 视觉提示注入护栏的证据对齐评测

    提出 Mind2Web-Injection 评测网页 Agent 视觉注入护栏

    arXiv
    2609.05535
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    GPT-5.6-luna, Qwen3-VL-32B, Qwen3-VL-8B and others6
    Abstract作者称检测、定位和反事实响应在六个 VLM 上可分开。

    Mind2Web-Injection 用来检查网页截图护栏的判决是否对齐到该用的画面证据。。

    Paper analysisRead analysis
  17. Evaluation and benchmarksarXiv 2608.16824

    GEO-Flag：检测与测量面向生成式引擎优化的网页内容

    提出 GEOFlagBench 与 IPT，检测并测量生成式引擎优化网页

    arXiv
    2608.16824
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    ModernBERT-base, Qwen3-0.6B, Gemini-3.5-flash and others11
    AbstractGEOFlagBench 与 IPT 用于检测 GEO 页，并估计真实检索中的占比与引用可核验性。

    GEO-Flag 把页面级 GEO 检测做成可评测任务，并在发布的检索结果上估计被判为 GEO 的页面占比。

    Paper analysisRead analysis
  18. AttacksarXiv 2608.00718

    研究揭示多智能体 LLM 流水线的结构性漏洞

    提出多智能体流水线的边界攻击，使对抗内容作为可信输入跨层传播

    arXiv
    2608.00718
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    GPT-5-mini, Claude Sonnet 4.5, Kimi K2.5
    Abstract缺边界校验使对抗内容跨智能体传播，作者称这是架构属性而非模型能力。

    作者把多智能体 LLM 流水线的对抗问题定义为缺少边界校验，并用生产轨迹和受控实验检查脆弱性来自架构还是骨干模型。中间输出未经验证就传给下游。作者认为，一旦某级接受对抗内容，后续智能体会把它当作可信输入。

    Paper analysisRead analysis
  20. AttacksarXiv 2605.24421

    研究者提出日志型提示注入攻击，可抑制 SOC 中 LLM 分析助手的恶意日志识别

    提出日志基底提示注入，操纵 SOC 分析员的分类、摘要与处置建议

    arXiv
    2605.24421
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    gpt-4o-mini

    Abstract在 gpt-4o-mini 上，人格劫持与上下文操纵可改写 SOC 分析输出，直接覆盖在分类上无效。

    Paper analysisRead analysis
Load more