Skip to content

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

Oct 12 · Monday

All papers

Found 35 papers
Filter2

Attack type

Jailbreaks10Prompt injection12Poisoning and backdoors8Extraction and theft3Detection evasion3Denial of service4

Attack technique

Automated search5Encoding and obfuscation3Role-play and persuasion3Conversation structure manipulation4Reasoning-chain manipulation14Cross-modal payloads3Defense-targeted4Dormant triggers5

Defense type

Permissions and approval3

Impact

Harmful content8Unauthorized actions8Data and privacy leaks5Output manipulation7Availability and reliability5

System type

Models and APIs15AI Agent18

Attacker access

Black-box10

Attack surface / entry point

User input20Documents and messages3Tool responses4Tool and skill descriptions4Code repositories and dependencies4Inter-agent messages3Images, audio and video3

Layer

Prompt layer12Application layer19

System component

MCP and skills38Memory24RAG26Guardrails and judges55Monitors6Reasoning chains23Permissions and sandboxing12Fine-tuning and open weights29Vision60Audio13Video3Image generation22Embeddings9Multimodal96
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 232papers have no tags yet and are not included in filtered results.
System component：Reasoning chainsSystem component：Permissions and sandboxingClear all

Another 232papers do not have category tags yet and are not included in filtered results.

  1. AttacksarXiv 2610.12644

    Cancerous backdoor attack: progressive activation via invisible tokens and a stage controller.

    Proposes the Cancerous backdoor, which progressively corrupts reasoning using invisible stage tokens.

    arXiv
    2610.12644
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Evaluated models
    Llama-3.1-8B-Instruct, Qwen2.5-7B-Instruct, Mistral-7B-Instruct-v0.3
    AbstractHidden tokens plus a stage controller make the backdoor progressively corrupt reasoning upon triggering, and the authors believe that looking only at the answer will miss it.

    Cancerous Backdoor Attack changes LLM backdoors from one-time activation to multi-stage reasoning corruption that accumulates progressively with triggers.

    Paper analysisRead analysis
  2. 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
  4. AttacksarXiv 2609.35699

    论文：蒸馏防御在蒸馏后强化学习下易被攻破

    提出用 API 推理摘要重建轨迹再蒸馏加强化学习，攻破蒸馏防御

    arXiv
    2609.35699
    Published
    Layer
    Training and data layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    Qwen2.5-0.5B, Qwen2.5-1.5B, Qwen2.5-3B and others12
    Abstract更现实的蒸馏攻击是蒸馏后再 RL。

    这篇工作修订蒸馏攻击的威胁模型：攻击者经合法 API 收集闭源推理数据，先蒸馏自有模型，再做强化学习。作者认为只在蒸馏后评估防御，会漏掉后续 RL，从而高估防御。

    Paper analysisRead analysis
  5. AttacksarXiv 2609.34518

    SEAD 论文提出工具型 Agent 的状态视角攻击与防御方法 DART 与 SAGE

    提出工具型 Agent 的自适应攻击 DART 与预执行防御 SAGE

    arXiv
    2609.34518
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    GPT-5.6 Luna, GPT-5.6 Terra, Gemini 3.8 Flash and others5

    Abstract论文提出状态控制框架 SEAD、攻击 DART 与防御 SAGE，实验显示执行反馈与环境探测在攻防交互中具有关键作用。

    Paper analysisRead analysis
  8. 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
  9. AttacksarXiv 2609.15989

    Researchers propose plan injection attack that can bypass chain-of-thought monitoring

    提出 plan injection 攻击，植入良性伪装计划以绕过思维链监视器

    arXiv
    2609.15989
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Gray-box
    Evaluated models
    Qwen2.5-Coder-7B-Instruct, Qwen3-8B, DeepSeek-R1-0528 and others6

    Abstract论文指出模型会将外部上下文中的伪装计划改写为自身推理，使思维链监视器失效甚至反向合理化恶意行为。

    Paper analysisRead analysis
  10. AttacksarXiv 2609.14003

    FlowSeal：用信息流控制阻断 LLM Agent 的隐私泄露

    提出 FLOWSEAL，用工具级信息流控制阻断智能体隐私泄露

    arXiv
    2609.14003
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    DeepSeek-V3.2, GPT-4.1-mini, Gemini-Flash-2.5

    Abstract作者分析大模型智能体隐私检查与攻击面重合问题，提出外部拦截防御 FLOWSEAL 并验证有效性。

    Paper analysisRead analysis
  11. AttacksarXiv 2609.11757

    Researchers disclose a Whisper prompt injection attack on the AP2 payment protocol and propose the A-VIP defense

    提出 AP2 支付协议的 Whisper 攻击与 A-VIP 绑定防御

    arXiv
    2609.11757
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    gemini-2.5-flash-lite, gemini-3.1-flash-lite, gemini-3-flash-preview and others15

    Abstract论文揭示了 AP2 协议决策层受操纵的风险，提出结构化与权能绑定防御 A-VIP，并发布评测基准。

    Paper analysisRead analysis
  12. AttacksarXiv 2609.09553

    研究：任意密文攻击前沿大模型无需微调即可越狱

    提出任意字母置换密码越狱，无需微调即可诱导有害输出

    arXiv
    2609.09553
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    Claude Sonnet 4, Claude Sonnet 4.5, Gemini 3 Flash (preview) and others7

    Abstract前沿大模型无需微调即可通过提示词学会字母置换密码，进而绕过分类器与对齐，作者呼吁关注能力提升带来的安全风险。

    Paper analysisRead analysis
  13. AttacksarXiv 2609.08186

    论文揭示大推理模型的对齐崩溃并提出 Reasoning Trap 越狱范式

    揭示推理加深引发的对齐坍塌，用 Reasoning Trap 放大越狱

    arXiv
    2609.08186
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-8B, Qwen3-14B, Llama2-7B and others8
    Abstract更深推理提升题目准确率，同时提高扰动下的相对损失。

    作者研究扩展推理是否以对齐稳健性为代价，提出 ALR、RT 与 RRA。。

    Paper analysisRead analysis
  15. AttacksarXiv 2609.01023

    Akrasia：针对推理型代码 LLM 的隐蔽后门攻击

    提出隐蔽后门攻击 AKRASIA，在推理代码模型提示中植入触发器并伪装思维链

    arXiv
    2609.01023
    Published
    Layer
    Prompt layer
    Setting
    coding agent
    Evaluated models
    Claude-Sonnet-5, GPT-5.5, Gemini-3.5-flash and others7

    Abstract论文针对推理代码大模型提出了隐蔽推断期后门攻击 AKRASIA，揭示了模型思维链可被利用进行欺骗性伪装的安全隐患。

    Paper analysisRead analysis
  16. AttacksarXiv 2608.29381

    Researchers propose five categories of safety failures in agent checkpoint rollback and implement three attacks on Hermes, Cline, and LangGraph

    提出利用检查点回滚破坏 Agent 执行连续性的攻击

    arXiv
    2608.29381
    Published
    Layer
    Application layer
    Setting
    AI Agent

    Abstract论文系统研究了智能体 C/R 的安全性，分析了合法回滚破坏执行连续性的机理，并实证评估了其普遍性与危害。

    Paper analysisRead analysis
  18. AttacksarXiv 2608.09867

    Researchers steal proprietary LLM inference chains via cross-model compatibility of encrypted inference blocks

    利用加密思维块的跨模型重放，诱导弱模型逐字转写专有推理链

    arXiv
    2608.09867
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5 and others13

    Abstract论文揭示了客户端加密思维块跨模型互换漏洞，提出了低成本提取方法并给出防御方案。

    Paper analysisRead analysis
  20. AttacksarXiv 2607.00422

    KidnapRAG：黑盒投毒攻击劫持 Agentic RAG 的推理链

    提出 KidnapRAG，用投毒文档黑盒劫持 Agentic RAG 的推理链

    arXiv
    2607.00422
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Qwen2.5-32B-Instruct, Llama-3.3-70B-Instruct, QwQ-32B and others5
    Abstract顺序投毒可把多步推理拖向指定应答，作者称关键在持续引导而非单次注入。

    KidnapRAG 是针对 Agentic RAG 的 black-box 语料投毒：攻击者不能访问模型、提示词、检索器或私有轨迹，只能发布可能被检索的文档，并利用公开界面上的推理与搜索查询。

    Paper analysisRead analysis
Load more