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

Oct 12 · Monday

All papers

Found 68 papers

Filter3

System type

Models and APIs6 LLM applications3 AI Agent60 coding agent14 web agent7 Browser and computer use5 Multi-agent systems3

Attack surface / entry point

User input10 Web pages and interfaces17 Documents and messages12 Tool responses17 Tool and skill descriptions38 Memory writes5 Inter-agent messages5 Images, audio and video18 Distribution channels24

Layer

Prompt layer5 Application layer62

System component

MCP and skills47 Memory35 RAG10 Guardrails and judges34 Monitors22 Reasoning chains7 Permissions and sandboxing45 Vision21 Multimodal22

Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 265papers have no tags yet and are not included in filtered results.

Attack type：Prompt injection System component：MCP and skills System component：Multimodal Clear all

Another 265papers do not have category tags yet and are not included in filtered results.

DefensesarXiv 2610.11942Oct 8
LGWM: Verifying GUI agent screen transitions with action-conditioned world models
Proposes LGWM, using an action-conditioned world model to verify GUI agent screen transitions.
arXiv
2610.11942
Published
Oct 8
Layer
Application layer
Setting
Browser and computer use
Evaluated models
LGWM
- DefensesMonitoring and auditing
- AttacksPrompt injection
- ComponentMonitors
Paper analysisRead analysis
AttacksarXiv 2610.12524Oct 8
An exploration of attacks on AI agent systems
重测并扩展对 AI Agent 的提示注入，并用 Agent Card 欺骗操纵 A2A 选择
arXiv
2610.12524
Published
Oct 8
Layer
Application layer
Setting
AI Agent
Evaluated models
Ollama Mistral 0.9.5, llama3-8b-8192, gemini-2.0-flash-lite and others5
- AttacksPrompt injection
- TechniqueEncoding and obfuscation
- ComponentMCP and skills
Abstract单智能体过滤与分类可被格式或措辞绕过，A2A 选卡还会偏向名称和描述。
作者在自建的 LangChain、AutoGen 和 A2A 系统上重测提示注入，并改 Agent Card 看选择是否偏移。摘要称试验主要凸显单智能体弱点，同时把攻击扩到多智能体。
Paper analysisRead analysis
DefensesarXiv 2610.09469Oct 7
Secure-CUA: Constraining the Impact of Untrusted Content in Computer-Use Agents
提出 Secure-CUA，用动作事务约束计算机使用智能体的不可信影响
arXiv
2610.09469
Published
Oct 7
Layer
Application layer
Setting
web agent
Evaluated models
Claude Opus 5, Gemini 3.8 Flash, GPT-5.6-Sol
- DefensesInstruction–data separation
- AttacksPrompt injection
- ComponentVision
AbstractSecure-CUA 用每步事务约束不可信影响，良性 TSR 接近 Vanilla-CUA。
Secure-CUA 要限制图形界面里不可信内容对 CUA 决策和 GUI 执行的影响，同时保留合法任务所需的不可信信息。
Paper analysisRead analysis
AttacksarXiv 2610.09240Oct 7
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
Oct 7
Layer
Application layer
Setting
web agent
Evaluated models
LLaVA-v1.5-13B, LLaVA-v1.6-34B, MiniCPM-o-8B and others6
- AttacksPrompt injection
- TechniqueGradient and representation optimization
- ComponentVision
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
AttacksarXiv 2610.09027Oct 7
P-VMI：对抗图像经 KV cache 持续影响多轮对话
提出持久视觉记忆注入，使对抗图像经缓存在触发时操控输出
arXiv
2610.09027
Published
Oct 7
Layer
Infrastructure layer
Setting
Models and APIs
Evaluated models
Qwen3-8B, Qwen3-VL-8B-Instruct, Gemma-4-12B
- AttacksPrompt injection
- TechniqueDormant triggers
- ComponentVision
Abstract论文提出 P-VMI，证实对抗输入被掩蔽后仍可经 KV 缓存持久控制模型，强调了保留状态的安全风险。
Paper analysisRead analysis
AttacksarXiv 2610.08098Oct 6
CORSA：面向技能路由器的技能注入攻击可跨路由器与 LLM 迁移
提出 CORSA，按任务簇优化技能注入的检索与执行
arXiv
2610.08098
Published
Oct 6
Layer
Application layer
Setting
coding agent
Evaluated models
GPT-5.4, DeepSeek-V4-Pro, Qwen3.8-27B and others5
- AttacksPoisoning and backdoors
- TechniqueSelection manipulation
- ComponentMCP and skills
Paper analysisRead analysis
Analysis and theoryarXiv 2610.06001Oct 5
AgentSpy: Observing AI agent behavior at the system call layer and detecting skill injection attacks
提出 AgentSpy，在系统调用层观测 Agent 行为并检测技能注入
arXiv
2610.06001
Published
Oct 5
Layer
Application layer
Setting
coding agent
Evaluated models
gpt-5.6-terra, deepseek-v4-flash, glm-5.3-flash and others4
- DefensesDetection and filtering
- AttacksPoisoning and backdoors
- ComponentMCP and skills
Abstract提出了系统级智能体监控框架 AgentSpy，揭示出通过测试背后的非预期行为与注入风险。
本研究提出了面向 AI 智能体的系统级外部监控与行为分析框架 AgentSpy。
Paper analysisRead analysis
DefensesarXiv 2610.06966Oct 4
APEX: proactively defending against indirect prompt injection at the LLM agent execution boundary.
提出 APEX，在执行边界用授权契约防御 Agent 间接提示注入
arXiv
2610.06966
Published
Oct 4
Layer
Application layer
Setting
AI Agent
Evaluated models
DeepSeek-V4-Flash, GLM-5.2, GPT-5.6-sol
- DefensesInstruction–data separation
- AttacksPrompt injection
- ComponentPermissions and sandboxing
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2610.04378Oct 3
COPEX：面向 MCP 智能体的受控攻击评测基准发布
提出 COPEX 基准，评测 MCP 智能体对多入口攻击的抵抗力
arXiv
2610.04378
Published
Oct 3
Layer
Application layer
Setting
AI Agent
Evaluated models
Claude Opus 4.7, Claude Sonnet 4.6, GPT-5.5 and others9
- AttacksPrompt injection
- ComponentMCP and skills
Paper analysisRead analysis
DefensesarXiv 2610.02861Oct 2
论文提出在 Kubernetes 上保护 AI Agent 的七层纵深防御框架与参考架构
提出 Kubernetes 运维智能体的七层纵深防御框架与参考架构
arXiv
2610.02861
Published
Oct 2
Layer
Application layer
Setting
AI Agent
- DefensesPermissions and approval
- AttacksPrompt injection
- ComponentPermissions and sandboxing
Abstract把运维智能体当不可信租户，用七层控制在完全攻破假设下约束行动。
把 Kubernetes 上的运维智能体当作不可信租户，在假定提示注入已完全控制其工具选择的前提下，用基础设施重建数据面与控制面的边界。要解决的是：智能体读日志、注解、工单或 MCP 工具描述时，这些内容可以变成它带着自身凭据执行的命令。模型对齐和分类器只降低概率。
Paper analysisRead analysis
DefensesarXiv 2610.02432Oct 2
论文提出 LLM 输入扰动鲁棒性度量，并发布 ASA 攻击与 AttestMCP 防护
提出 ASA 与 AttestMCP，度量并降低 LLM 对输入序列扰动的敏感
arXiv
2610.02432
Published
Oct 2
Layer
Application layer
Setting
LLM applications, AI Agent
Attacker
Black-box
Evaluated models
Gemma-3-27B-Instruct, Gemma-3-4B-Instruct, Llama-3.2-3B-Instruct and others12
- DefensesDetection and filtering
- AttacksDetection evasion
- ComponentGuardrails and judges
Paper analysisRead analysis
DefensesarXiv 2610.03089Oct 2
COBRA：为计算机使用 Agent 防御分支引导攻击的双 LLM 架构
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
Oct 2
Layer
Application layer
Setting
Browser and computer use
Evaluated models
gpt-5, UI-TARS-1.5-7B, claude-sonnet-5 and others6
- DefensesPermissions and approval
- AttacksPrompt injection
- ComponentPermissions and sandboxing
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
Evaluation and benchmarksarXiv 2610.03153Oct 2
EvoRiskBench：面向工作区 Agent 运行时安全风险的演化基准
构建 EvoRiskBench 评测工作区 Agent 的运行时安全风险
arXiv
2610.03153
Published
Oct 2
Layer
Application layer
Setting
AI Agent
Evaluated models
GPT-5.6 Sol, DeepSeek-V4-Pro-0813, Claude Opus 5
- AttacksPrompt injection
- ComponentMCP and skills
Abstract论文构建了工作空间智能体运行时风险基准 EvoRiskBench，发现高执行力智能体在 IPI 下面临较高安全风险。
Paper analysisRead analysis
AttacksarXiv 2610.01564Oct 1
APEX 用跨技能链劫持 LLM Agent
提出 APEX，用跨技能进度记录夹带虚假授权劫持 Agent
arXiv
2610.01564
Published
Oct 1
Layer
Application layer
Setting
AI Agent
Evaluated models
Claude Sonnet 5, DeepSeek V4 Flash, GPT-5.4 and others6
- AttacksPrompt injection
- TechniqueAutomated search
- ComponentMCP and skills
Paper analysisRead analysis
DefensesarXiv 2610.01349Oct 1
PACE：面向工具调用 LLM Agent 的来源感知能力强制执行
提出 PACE，在工具调用前做路径隔离与效果授权
arXiv
2610.01349
Published
Oct 1
Layer
Application layer
Setting
AI Agent
Evaluated models
gpt-5.6-luna, DeepSeek-V4-Flash, Qwen-3.8-27B
- DefensesInstruction–data separation
- AttacksPrompt injection
- ComponentPermissions and sandboxing
AbstractPACE 在工具边界做路径割与效果核验。
PACE 是面向工具调用型 LLM 智能体的运行时调解：当准入证据无法区分会改变行为的规格时，不在出口站点上撤销调解，而在最后一次工具调用前同时检查表示出的影响路径和请求权限。
Paper analysisRead analysis
AttacksarXiv 2610.00392Sep 30
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
Sep 30
Layer
Application layer
Setting
Multi-agent systems
Evaluated models
MiniMax-M3, deepseek-v4-pro, glm-5.3 and others8
- AttacksPrompt injection
- TechniqueMulti-turn escalation
- ComponentMCP and skills
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
DefensesarXiv 2609.39450Sep 30
ActionGuard：在工具调用执行前拦截被投毒技能
提出 ActionGuard 在执行前拦截被投毒技能诱发的未授权工具调用
arXiv
2609.39450
Published
Sep 30
Layer
Application layer
Setting
AI Agent
Evaluated models
Gemma 4 (gemma4:e4b-it-q4_K_M), Qwen 3.5 (qwen3.5:9b), Ministral 3 (ministral-3:8b-instruct-2512-q4_K_M) and others5
- DefensesInstruction–data separation
- AttacksPoisoning and backdoors
- ComponentMCP and skills
Paper analysisRead analysis
AttacksarXiv 2609.39065Sep 30
TrustProbe 在 11 个开源 Agent 中发现 104 个技能信任链漏洞
提出 TrustProbe，挖掘技能智能体中不可信 skill 到特权操作的信任链漏洞
arXiv
2609.39065
Published
Sep 30
Layer
Application layer
Setting
AI Agent
Evaluated models
OpenClaw, OpenCode, Hermes Agent and others11
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentMCP and skills
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2610.06898Sep 29
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
Sep 29
Layer
Application layer
Setting
Browser and computer use
Evaluated models
Qwen2.5-VL(32B), Qwen3-VL(32B), GUI-Owl(32B) and others7
- AttacksPrompt injection
- ComponentVision
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
DefensesarXiv 2609.36121Sep 29
Render Before Reading：把不可信内容渲染成图像以防御提示注入
提出 Pictionary，将不可信文本渲染成图像以防御提示注入
arXiv
2609.36121
Published
Sep 29
Layer
Application layer
Setting
AI Agent
Evaluated models
Claude Opus 4.7, Claude Haiku 4.5, GPT-5.5 and others12
- DefensesInstruction–data separation
- AttacksPrompt injection
- ComponentVision
Paper analysisRead analysis

AI Safety Daily

Read the latest issueOne issue per day, updated as events develop

All papers

Attack type

Attack technique

Defense type

Impact

System type

Attack surface / entry point

Layer

System component

LGWM: Verifying GUI agent screen transitions with action-conditioned world models

An exploration of attacks on AI agent systems

Secure-CUA: Constraining the Impact of Untrusted Content in Computer-Use Agents

WebMirage hijacks visual web agents' browser operations with adversarial images

P-VMI：对抗图像经 KV cache 持续影响多轮对话

CORSA：面向技能路由器的技能注入攻击可跨路由器与 LLM 迁移

AgentSpy: Observing AI agent behavior at the system call layer and detecting skill injection attacks

APEX: proactively defending against indirect prompt injection at the LLM agent execution boundary.

COPEX：面向 MCP 智能体的受控攻击评测基准发布

论文提出在 Kubernetes 上保护 AI Agent 的七层纵深防御框架与参考架构

论文提出 LLM 输入扰动鲁棒性度量，并发布 ASA 攻击与 AttestMCP 防护

COBRA：为计算机使用 Agent 防御分支引导攻击的双 LLM 架构

EvoRiskBench：面向工作区 Agent 运行时安全风险的演化基准

APEX 用跨技能链劫持 LLM Agent

PACE：面向工具调用 LLM Agent 的来源感知能力强制执行

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

ActionGuard：在工具调用执行前拦截被投毒技能

TrustProbe 在 11 个开源 Agent 中发现 104 个技能信任链漏洞

DIBench：面向移动 GUI 智能体多候选选择决策完整性的安全基准

Render Before Reading：把不可信内容渲染成图像以防御提示注入