Skip to content

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

Oct 13 · Tuesday

All papers

Found 96 papers
Filter5

Attack type

Jailbreaks51Prompt injection17Poisoning and backdoors6Extraction and theft5Detection evasion10

Attack technique

Gradient and representation optimization54Automated search113Multi-turn escalation30Encoding and obfuscation68Role-play and persuasion34Conversation structure manipulation14Reasoning-chain manipulation15Cross-modal payloads61Defense-targeted28Dormant triggers39Selection manipulation12

Defense type

Detection and filtering4

Impact

Harmful content48Unauthorized actions6Data and privacy leaks5Output manipulation33Availability and reliability3

System type

Models and APIs64LLM applications11AI Agent21web agent4Multi-agent systems3Embodied systems and robotics4

Attacker access

Black-box34Gray-box3White-box16

Attack surface / entry point

User input56Web pages and interfaces5Documents and messages5Tool responses6Retrieval corpus5Images, audio and video35

Layer

Prompt layer67Model layer13Application layer29Training and data layer3

System component

RAG4Guardrails and judges9Reasoning chains5Vision32Audio5Embeddings3Multimodal38
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 203papers have no tags yet and are not included in filtered results.
Attack technique：Gradient and representation optimizationAttack technique：Multi-turn escalationAttack technique：Conversation structure manipulationLayer：Prompt layerLayer：Application layerClear all

Another 203papers do not have category tags yet and are not included in filtered results.

  3. AttacksarXiv 2610.09240

    WebMirage hijacks visual web agents' browser operations with adversarial images

    提出 WebMirage，用局部对抗图像劫持视觉网页 Agent 的浏览器操作

    arXiv
    2610.09240
    Published
    Layer
    Application layer
    Setting
    web agent
    Evaluated models
    LLaVA-v1.5-13B, LLaVA-v1.6-34B, MiniCPM-o-8B and others6
    Abstract论文提出了针对网络智能体的端到端视觉红队框架 WebMirage，揭示了从视觉定位到浏览器执行的安全脆弱性。

    WebMirage 是一个面向视觉定位网络智能体的流水线感知端到端红队攻击框架。

    Paper analysisRead analysis
  6. AttacksarXiv 2610.04985

    研究揭示 Jev 作为应用决策层的安全与隐私风险

    评估应用决策层的输入注入与隐私推断风险

    arXiv
    2610.04985
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Attacker
    Black-box, Gray-box
    Evaluated models
    Jev, NanoJev
    AbstractJev 的类型化决策可被输入与训练操纵，也会泄露信息，并同时支持检测与滥用。

    Jev 是 TypeSafe 的 System One 模型，把应用 state 和固定问题变成 Choice、Noul 或 Score 的类型化决策与概率。作者用官方 API 和可改训练的 NanoJev，考察它作为决策层时的操纵、泄露与双重用途。

    Paper analysisRead analysis
  8. 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
  9. AttacksarXiv 2610.00392

    A2A-TIBA：针对 A2A 协议 LLM 智能体的间接提示注入攻击与三层同构攻防模型

    提出 A2A-TIBA，经 A2A 对等任务间接注入并植入常驻回调

    arXiv
    2610.00392
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems
    Evaluated models
    MiniMax-M3, deepseek-v4-pro, glm-5.3 and others8
    AbstractA2A-TIBA 一次植入后绕过智能体，GDA 用四类结果定位防线，ELA-ITL 把信封标注列为第三层。

    A2A-TIBA 把间接提示注入和绕过放进 A2A：gray-box 攻击者已能投递任务，诱导目标部署常驻回调程序后，下令和外传不再经过智能体。

    Paper analysisRead analysis
  10. AttacksarXiv 2609.36956

    The researchers propose Controlled Decoding Attacks, jailbreaking black-box LLMs through a text-only interface

    提出 BLINDBIAS，通过纯文本续写与前缀控制越狱黑盒模型

    arXiv
    2609.36956
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    GLM-5, Gemini-3.5-Flash, Qwen3-32B and others4
    Abstract作者认为，隐藏数值概率并不足以消除支持重复采样和前缀续写的解码攻击面。

    BLINDBIAS 是一种针对 sample-only 续写接口的 black-box 解码时越狱框架，研究没有目标数值概率时，采样能否支持逐步控制。

    Paper analysisRead analysis
  11. Evaluation and benchmarksarXiv 2609.36849

    Study evaluates GradSafe's jailbreak detection vulnerability in multi-turn dialogues

    评测多轮对话中 GradSafe 越狱检测的可分性与脆弱性

    arXiv
    2609.36849
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Llama 3.1 8B Instruct, Qwen2.5-7B-Instruct
    AbstractCWS 显示不安全梯度在真实多轮流量中变弱，且随攻击族和模型而变。

    作者评估梯度越狱检测从单条提示词迁到多轮对话后还剩多少可分性，方法是用 Context Window Scanner 把 GradSafe 原封不动地套到用户轮窗口上。

    Paper analysisRead analysis
  13. AttacksarXiv 2609.27542

    Researchers use control token injection to delete gpt-oss-20b's reasoning chain and bypass chain-of-thought monitoring

    提出控制标记注入，清空工具 Agent 的思维链以绕过推理监控

    arXiv
    2609.27542
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    openai/gpt-oss-20b, google/gemma-4-26B-A4B-it, deepseek-ai/DeepSeek-R1-Distill-Qwen-7B and others4

    Abstract论文阐明控制标记注入可抑制 CoT 使监控失效并促成拒绝绕过，且工具执行受解析器宽容度影响。

    Paper analysisRead analysis
  14. InterpretabilityarXiv 2609.24801

    研究用 XAI 归因分析 Prompt Guard 2，同义词替换可翻转其提示注入判定

    用词级归因指导同义词替换，翻转提示注入检测并验证越狱

    arXiv
    2609.24801
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Prompt Guard 2 (86M), Llama 3.1 8B

    Abstract作者称检测靠分散的词法线索。

    Paper analysisRead analysis
  15. AttacksarXiv 2609.15383

    微软团队提出 capability laundering 攻击：弱模型拆分任务向对齐前沿模型套取能力

    提出能力洗钱攻击，拆分任务向对齐模型套取能力

    arXiv
    2609.15383
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Attacker
    Black-box
    Evaluated models
    GPT-5.5, Claude Opus 4.8, Grok-4.3 and others7

    Abstract作者提出了能力洗钱攻击，发现未对齐编排器分解咨询对齐模型可恢复能力差距，指出需建立跨会话防御。

    Paper analysisRead analysis
  17. AttacksarXiv 2609.03247

    研究揭示大语言模型的自发身份认证失败：五款模型的分阶段开发者身份实验

    提出 MIPC 与 CFA，观察模型自拟测验后接受开发者身份

    arXiv
    2609.03247
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    ChatGPT, Claude, Qwen3.7-Plus and others5
    Abstract自拟测验可造成对话层虚假认证，但所测授权边界并未因此改变。

    这项工作把“模型自己出题、自己批改、再宣布身份已核验”定义成对话层认证失败，并与后端提权分开。

    Paper analysisRead analysis
  18. AttacksarXiv 2609.02414

    Blueprint 框架用世界观模拟与 18 个影响因子提升多轮越狱成功率

    提出多轮越狱框架 BLUEPRINT，用世界观模拟与心理因素诱导有害输出

    arXiv
    2609.02414
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    Qwen3-Next-80B, DeepSeek-V3.2, GLM-4.7 and others6
    AbstractBLUEPRINT 解耦多轮越狱策略与情境模拟，揭示任务具象化是摆脱拒绝的核心机制，为多轮安全防御提供新视角。

    BLUEPRINT 是一个将因子化社会影响力策略空间与跨轮世界观模拟解耦的多轮大语言模型安全评测框架。

    Paper analysisRead analysis
  19. AttacksarXiv 2609.02172

    BOSS：以广度优先后缀搜索改进 GCG 越狱优化

    提出广度优先后缀搜索 BOSS，改进 GCG 越狱优化

    arXiv
    2609.02172
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    White-box
    Evaluated models
    Llama-2-7B-Chat, Yi-1.5-9B-Chat, Qwen2-7B-Instruct and others8
    AbstractBOSS 把 GCG 类优化从单条深搜索改为多条短轨迹加源侧诊断。

    BOSS 是插在 GCG 类离散 suffix 优化上的搜索框架：局部梯度更新不变，变的是固定预算花在深度还是广度。

    Paper analysisRead analysis
Load more