Skip to content

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

Oct 13 · Tuesday

All papers

Found 134 papers
Filter6

Attack type

Jailbreaks67Prompt injection40Poisoning and backdoors42Extraction and theft3Detection evasion9

Attack technique

Gradient and representation optimization65Automated search115Multi-turn escalation30Encoding and obfuscation67Role-play and persuasion34Conversation structure manipulation15Reasoning-chain manipulation16Cross-modal payloads61Defense-targeted35Dormant triggers41Selection manipulation12

Defense type

Detection and filtering7

Impact

Harmful content64Unauthorized actions22Data and privacy leaks14Output manipulation29Oversight failure3

System type

Models and APIs70LLM applications11AI Agent53coding agent14Multi-agent systems5

Attacker access

Black-box45

Attack surface / entry point

User input91Web pages and interfaces8Documents and messages12Tool responses12Tool and skill descriptions12Code repositories and dependencies7Retrieval corpus6Memory writes19Inter-agent messages4Images, audio and video23Weights and gradients5Distribution channels14

Layer

Prompt layer68Model layer3Application layer63Training and data layer22

System component

MCP and skills14Memory26RAG7Guardrails and judges26Reasoning chains4Permissions and sandboxing5Fine-tuning and open weights3Vision20Audio3Image generation7Multimodal28
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 215papers have no tags yet and are not included in filtered results.
Attack technique：Multi-turn escalationAttack technique：Encoding and obfuscationAttack technique：Dormant triggersLayer：Prompt layerLayer：Model layerLayer：Application layerClear all

Another 215papers do not have category tags yet and are not included in filtered results.

  1. AttacksarXiv 2610.13928

    AI models can provably hide arbitrary capabilities

    Proposes digital lock compilation, encrypting executable networks into host weights to hide arbitrary capabilities

    arXiv
    2610.13928
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Gemma 3 270M
    AbstractThe authors use a digital lock to hide an executable capability and state that elicitation on the chemistry and parity tasks did not recover the capability.

    The authors construct a class of backdoors that encrypt an executable network into the weights, used to discuss the limits of white-box elicitation audits under supply chain attacks.

    Paper analysisRead analysis
  2. Evaluation and benchmarksarXiv 2610.11766

    研究提出意图恢复率作为 LLM 安全评测的补充指标

    提出 IRIS，用意图恢复率区分遮蔽越狱下无害输出的原因

    arXiv
    2610.11766
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    GPT-4o, Claude Sonnet 4.5, Claude Opus 4.5 and others6
    Abstract无害输出的证据权重取决于任务是否被恢复。

    IRIS 把意图恢复当作中间信号，用来解释意图遮蔽提示下为何没有出现有害协助。

    Paper analysisRead analysis
  3. AttacksarXiv 2610.12524

    An exploration of attacks on AI agent systems

    Retest and extend prompt injection against AI agents, and use Agent Card deception to manipulate A2A selection

    arXiv
    2610.12524
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Ollama Mistral 0.9.5, llama3-8b-8192, gemini-2.0-flash-lite and others5
    AbstractSingle-agent filtering and classification can be bypassed by format or wording, and A2A card selection is also biased toward names and descriptions.

    The authors retested prompt injection on their own LangChain, AutoGen, and A2A systems, and modified the Agent Card to see if choices shifted. The abstract states the experiments mainly highlight single-agent weaknesses while extending attacks to multi-agent.

    Paper analysisRead analysis
  4. Evaluation and benchmarksarXiv 2610.09033

    研究发布 ASRD 数据集，评测五款开源模型在非规范输入下的安全表现

    arXiv
    2610.09033
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Gemma 2 9B, Gemma 3 4B, Llama 3.1 8B and others6
    Abstract五模型、七族表面形式的四态评测里，低有害遵从不等于读懂后的安全处理。

    ASRD 用四态标签检查五个开源模型如何处理非规范表面上的有害请求。

    Paper analysisRead analysis
  5. AttacksarXiv 2610.07645

    SkillPoison：用成功经验投毒自进化 Agent 的技能形成

    提出 SkillPoison，用局部有效的成功经验诱导自进化 Agent 形成过度泛化技能

    arXiv
    2610.07645
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    deepseek-v4-flash, Qwen V3.8 Flash, GPT-5.5
    AbstractSkillPoison 通过局部有效且验证成功的经验诱导技能过度泛化，在多个评测基准上实现了持久的下游攻击影响。

    SkillPoison 提出了一种针对自进化智能体经验到技能形成过程的攻击范式，通过组织已验证的成功经验诱导技能过度泛化。

    Paper analysisRead analysis
  6. AttacksarXiv 2610.07510

    PersistBD: attackers can strengthen backdoors before release so that they maintain high attack success rates even after developer SFT and RL.

    提出 PERSISTBD，发布前强化供应链后门使其经良性 SFT 与 RL 后仍外发凭据

    arXiv
    2610.07510
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Qwen2.5-Coder-3B-Instruct, Qwen2.5-Coder-7B-Instruct, Qwen3-Coder-30B-A3B-Instruct
    Abstract总结了论文关于后门在 SFT/RL 中存活的发现、两因子机制、PERSISTBD 方法与核心数字。

    本文研究第三方 LLM 智能体在开发者进行良性监督微调（SFT）和强化学习（RL）后训练时的后门持久性风险。

    Paper analysisRead analysis
  7. AttacksarXiv 2610.06049

    研究者提出仅修改 SAE 解码器的后门攻击，可在不改动语言模型的情况下植入行为

    提出只改 SAE 解码器的后门，在冻结语言模型上植入代码插入

    arXiv
    2610.06049
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Gemma 3 4B Instruct, Qwen3 1.7B, Llama 3.1 8B Instruct
    Abstract只改解码器的 SAE 可带上代码插入后门，且若干常规质量指标变化不大。

    只改 SAE 解码器、冻结编码器与语言模型，即可把代码插入行为放进推理时的辅助组件。作者把这视为 SAE 供应链问题：checkpoint 一旦替换某一层激活，就参与后续计算，而不只是解释工具。

    Paper analysisRead analysis
  8. AttacksarXiv 2610.04589

    StegoMemory red team testing: agent persistent memory can be used as a cross-session covert channel

    提出 StegoMemory，评测智能体持久记忆作为跨会话隐写信道

    arXiv
    2610.04589
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Llama 3.1 405B, Claude 3.5 Sonnet, GPT-4o and others13

    Abstract研究揭示智能体持久记忆可作为隐蔽信道跨会话传输机密，写入持久化是主要防御瓶颈，呼吁建立记忆完整性防御。

    Paper analysisRead analysis
  10. AttacksarXiv 2610.02302

    研究者提出意图隐藏越狱的信息论框架，分析组合式攻击

    提出组合式意图隐藏越狱，按先验后验匹配选择辅助任务

    arXiv
    2610.02302
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-1.7B, Qwen3-4B, Qwen3-8B and others7
    Abstract用信息论选含目标的任务组合，再看查询能否既隐藏意图又让目标被执行。

    这篇工作把组合式意图隐藏写成任务选择：有害目标必须留在 bundle 里，同时让估计的有害意图概率贴近全集先验，或降到阈值以下。

    Paper analysisRead analysis
  12. 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
  14. 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
  15. AttacksarXiv 2609.38253

    CollageAttack: Exploiting cross-modal alignment defects for spatial text composition jailbreaks against T2I models.

    提出 CollageAttack，用空间文本碎片重组越狱文生图模型

    arXiv
    2609.38253
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    GPT-Image-2, Doubao-Seedream-5.0-Lite, Wan2.7-Image-Pro and others5

    Abstract系统揭示 T2I 跨模态空间合成安全漏洞，证实离散无害元素可在图像层面重构有害意图并放大传播强度。

    Paper analysisRead analysis
  16. Evaluation and benchmarksarXiv 2609.36849

    研究评估 GradSafe 在多轮对话中的越狱检测脆弱性

    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
  18. AttacksarXiv 2609.35576

    研究提出跨 LLM 助手的记忆跳跃攻击

    提出跨独立 Agent 的记忆跳跃攻击，经共享工件在助手间自传播

    arXiv
    2609.35576
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    GPT-5.6 Luna, Kimi-K2.6, GPT-OSS-120B and others6
    Abstract论文证实持久工件与内存可形成跨智能体传播链，揭示了将共享工件纳入安全防线的重要意义。

    该论文系统研究并量化了通过共享持久工件在具有独立私有内存的 LLM 智能体之间发生自传播的“工件介导传播”攻击风险。

    Paper analysisRead analysis
  19. Evaluation and benchmarksarXiv 2609.35912

    MMSkillRisk：多模态技能图像中的隐藏指令可诱导 Agent 执行未授权操作

    提出 MMSkillRisk 基准，评测多模态技能图像中的隐藏指令诱导 Agent 越权操作

    arXiv
    2609.35912
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    GPT-5.6-sol, Kimi-K2.6, Kimi-K3 and others6

    Abstract论文提出 MMSkillRisk 基准与 NCVA 攻击，揭示多模态技能中教学图像潜藏指令可引发未授权操作并与良性任务共存。

    Paper analysisRead analysis
  20. AttacksarXiv 2609.37468

    论文提出针对 Agentic RAG 的检索器后门攻击与 DOU 隐藏方法

    提出检索器后门攻击操纵智能体搜索，并用 DOU 逃避检测

    arXiv
    2609.37468
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    E5, Contriever-MSMARCO

    Abstract论文展示了通过恶意检索器操纵智能体多轮搜索并利用诱饵重塑表征隐蔽后门的可行性与系统风险。

    Paper analysisRead analysis
Load more