Skip to content

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

Oct 12 · Monday

All papers

Found 18 papers
Filter3

Attack type

Prompt injection8Poisoning and backdoors6Detection evasion7

Attack technique

Gradient and representation optimization33Automated search20Encoding and obfuscation10Role-play and persuasion11Conversation structure manipulation4Cross-modal payloads17Defense-targeted14Dormant triggers24Selection manipulation5

Impact

Harmful content17Unauthorized actions10Data and privacy leaks7Output manipulation18Oversight failure4

System type

Models and APIs9LLM applications3AI Agent6

Attacker access

Black-box4

Attack surface / entry point

User input8Images, audio and video4Training data3Distribution channels4

Layer

Prompt layer3Model layer4Application layer8Training and data layer3

System component

Guardrails and judges4Fine-tuning and open weights4Multimodal3
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 261papers have no tags yet and are not included in filtered results.
Attack technique：Conversation structure manipulationAttack technique：Defense-targetedImpact：Output manipulationClear all

Another 261papers 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 2610.04985

    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
  4. AttacksarXiv 2610.03166

    LiBRA：通过双向隐空间优化实现检测感知的图像水印去除

    提出 LiBRA，在潜空间双向优化以规避图像水印双侧检测

    arXiv
    2610.03166
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Attacker
    White-box
    Evaluated models
    Stable Signature, YU1, HiDDeN and others4

    AbstractLiBRA 用双向潜空间优化规避双侧水印检测，并提高相对反转基线的保真。

    Paper analysisRead analysis
  5. 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
  6. AttacksarXiv 2609.30383

    研究者提出技能级联攻击

    提出技能级联攻击，把有害目标拆散到多个技能里绕过单技能扫描

    arXiv
    2609.30383
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    GPT-5.4, GPT-5.4 Mini, Claude Opus 4.6 and others11

    Abstract论文探讨了多技能智能体系统面临的技能级联攻击风险，提出自动化红队框架与基准集，指出跨技能交互存在安全防护盲区。

    Paper analysisRead analysis
  8. AttacksarXiv 2609.08236

    研究：内容不变的风格包装可翻转 LLM 安全评判器的判定

    提出内容不变样式包装，翻转安全评审器的判定

    arXiv
    2609.08236
    Published
    Layer
    Prompt layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    GPT-4o-mini (gpt4o), GPT-4o-mini (strongreject), DeepSeek-Chat and others7
    Abstract论文揭示特定安全评审器易受保持内容不变的外层包装诱导改变判定，并指出安全评估应将判定稳定性作为重要考量指标。

    这是一项关于大语言模型自动化安全评审系统表面样式脆弱性的评估研究。

    Paper analysisRead analysis
  12. 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
  14. AttacksarXiv 2607.14493

    LogInject 框架披露 LLM 日志分析中的被动提示注入

    提出 LogInject 框架，评测日志分析中的被动提示注入及缓解

    arXiv
    2607.14493
    Published
    Layer
    Application layer
    Setting
    LLM applications
    Evaluated models
    GPT-4o, Claude 3.5 Sonnet, Llama-3-70B-Instruct

    Abstract论文评估了 LLM 日志分析中的上下文污染风险，提出分片拼接等攻击与三层防御，指出必须结合架构隔离与人工介入。

    Paper analysisRead analysis
  15. 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
  17. AttacksarXiv 2603.24203

    研究者提出 TIP：用树结构搜索为 MCP 生成隐蔽注入载荷

    提出 TIP，用树搜索为 MCP 响应生成隐蔽注入载荷

    arXiv
    2603.24203
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Qwen2.5-7B-Instruct, Qwen2.5-72B-Instruct, Llama3.1-8B-Instruct and others8
    AbstractTIP 在 MCP 工具响应上做黑盒注入。

    TIP 是面向 MCP 工具响应的 black-box 间接提示注入，目标是在不改工具元数据的情况下生成语义连贯的 JSON 载荷。

    Paper analysisRead analysis
  18. AttacksarXiv 2601.04261

    针对大语言模型后门指纹的抑制攻击

    提出 TFA 与 SVA，抑制 ensemble 后门指纹验证

    arXiv
    2601.04261
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    LLaMA2-7B, LLaMA3.1-8B-It, Qwen2.5-7B and others11
    Abstract两种 ensemble 攻击 TFA/SVA 抑制后门指纹。

    TFA 与 SVA 是针对后门式 LLM 指纹的 ensemble 抑制攻击，不改模型参数。所有者用 SFT 写入秘密对，只通过 API 验证版权。攻击者未授权取得模型，不知道触发和指纹，却要把每个模型的验证变成失败，同时让 ensemble 至少不差于性能最高的单模型。

    Paper analysisRead analysis
You have reached the end