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

Oct 13 · Tuesday

All papers

Found 30 papers

Filter6

Attack type

Prompt injection15 Poisoning and backdoors19

Attack technique

Gradient and representation optimization13 Automated search27 Encoding and obfuscation14 Role-play and persuasion8 Conversation structure manipulation3 Cross-modal payloads9 Defense-targeted9 Dormant triggers15 Selection manipulation4

Impact

Unauthorized actions10 Data and privacy leaks5 Output manipulation13

System type

Models and APIs75 LLM applications13 AI Agent52 coding agent13 Multi-agent systems4 Embodied systems and robotics5

Attack surface / entry point

User input8 Documents and messages5 Tool and skill descriptions7 Code repositories and dependencies4 Retrieval corpus4 Memory writes3 Images, audio and video4 Training data4 Weights and gradients5 Distribution channels10

Layer

Application layer24 Training and data layer5

System component

MCP and skills8 Memory4 RAG3 Guardrails and judges5 Fine-tuning and open weights5

Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 227papers have no tags yet and are not included in filtered results.

Attack technique：Encoding and obfuscation Attack technique：Conversation structure manipulation Attack technique：Dormant triggers System type：LLM applications System type：coding agent System type：Embodied systems and robotics Clear all

Another 227papers do not have category tags yet and are not included in filtered results.

AttacksarXiv 2610.07510Oct 6
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
Oct 6
Layer
Application layer
Setting
coding agent
Evaluated models
Qwen2.5-Coder-3B-Instruct, Qwen2.5-Coder-7B-Instruct, Qwen3-Coder-30B-A3B-Instruct
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentFine-tuning and open weights
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
AttacksarXiv 2610.05793Oct 5
研究者训练出按多智能体拓扑触发的代码后门模型
提出拓扑条件后门，使代码模型在推断多智能体部署时植入漏洞
arXiv
2610.05793
Published
Oct 5
Layer
Training and data layer
Setting
coding agent
Evaluated models
Qwen2.5-7B-Instruct, Qwen2.5-Coder-7B-Instruct
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentFine-tuning and open weights
Abstract论文证实了拓扑条件后门的可训练性与泛化性，展示了常规评测盲区，并提出差异化拓扑审计作为检测防御方向。
Paper analysisRead analysis
AttacksarXiv 2610.04985Oct 4
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
Oct 4
Layer
Application layer
Setting
LLM applications
Attacker
Black-box, Gray-box
Evaluated models
Jev, NanoJev
- AttacksPrompt injection
- TechniqueRole-play and persuasion
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
Evaluation and benchmarksarXiv 2609.35912Sep 28
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
Sep 28
Layer
Application layer
Setting
coding agent
Evaluated models
GPT-5.6-sol, Kimi-K2.6, Kimi-K3 and others6
- AttacksPrompt injection
- TechniqueCross-modal payloads
- ComponentMCP and skills
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
AttacksarXiv 2609.22711Sep 19
UBA-ORL：针对离线强化学习的遗忘激活后门攻击
提出 UBA-ORL，以 BD/CM 样本在遗忘后激活离线 RL 后门
arXiv
2609.22711
Published
Sep 19
Layer
Training and data layer
Setting
Embodied systems and robotics
Evaluated models
TD3+BC, BCQ, IQL
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
AbstractUBA-ORL 用竞争的 BD/CM 样本，使离线 RL 后门在删除伪装轨迹后激活。
UBA-ORL 研究离线强化学习中由合规轨迹删除激活的后门：合法贡献者同时上传共享触发的后门轨迹和伪装轨迹，训练后触发响应被压低，删除伪装批次后再升高。受害系统是使用静态数据集和轨迹级遗忘接口的离线 RL 服务。对手不能看到训练算法、参数、梯度或具体遗忘算法，但能提交可区分批次，并在部署时向观测注入预定触发。
Paper analysisRead analysis
AttacksarXiv 2609.18217Sep 16
研究：跨通道碎片化攻击可绕过 MCP 单通道防御
提出跨通道分段攻击，利用 MCP 隐式信任外发敏感文件
arXiv
2609.18217
Published
Sep 16
Layer
Application layer
Setting
coding agent
Evaluated models
GPT-4o, GPT-4o-mini, GPT-5.4 and others15
- AttacksPrompt injection
- TechniqueEncoding and obfuscation
- ComponentMCP and skills
Abstract论文揭示了 MCP 多通道隐式信任层级，通过跨通道分段与协议级利用突破现有防御，展示了严重的凭证外发风险。
Paper analysisRead analysis
AttacksarXiv 2609.13889Sep 12
PMPA：针对 OpenClaw 与 Claude Code 的持久记忆投毒攻击
提出 PMPA，经外部源向 Agent 持久记忆投毒并跨会话泄露隐私
arXiv
2609.13889
Published
Sep 12
Layer
Application layer
Setting
coding agent
Evaluated models
DeepSeek-V4-Flash, DeepSeek-V4-Pro, Qwen3-Max
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentMemory
Abstract论文提出了持久记忆投毒攻击 PMPA，报告了 harness 智能体在处理外部数据时易受跨会话记忆投毒与隐私泄露的风险。
Paper analysisRead analysis
AttacksarXiv 2609.07051Sep 7
TrojanWorld：通过想象引导对世界模型智能体植入后门
提出 TrojanWorld，用物理触发器对世界模型智能体植入想象引导后门
arXiv
2609.07051
Published
Sep 7
Layer
Training and data layer
Setting
Embodied systems and robotics
Evaluated models
TD-MPC2, DreamerV3, R2-Dreamer
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentFine-tuning and open weights
AbstractTrojanWorld 揭示了世界模型智能体的供应链后门风险，通过在世界模型中操纵想象可实现对闭环动作的持续控制。
Paper analysisRead analysis
AttacksarXiv 2609.02774Sep 3
CodePoisonRAG：针对检索增强代码生成的知识投毒攻击
提出 CodePoisonRAG，对检索增强代码生成的知识库投毒以植入漏洞
arXiv
2609.02774
Published
Sep 3
Layer
Application layer
Setting
LLM applications
Evaluated models
Qwen 3.5 9B, Code Llama 13B, DeepSeek-Coder-V2 16B
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentRAG
AbstractCodePoisonRAG 展示了 RACG 外部知识库的新攻击面，在较低投毒率下可定向诱发指定漏洞并部分抵御防御。
Paper analysisRead analysis
AttacksarXiv 2609.01222Sep 1
Researchers propose context privilege escalation attack, affecting 12 agent frameworks
提出针对 Agent Harness 的上下文特权提升攻击
arXiv
2609.01222
Published
Sep 1
Layer
Application layer
Setting
coding agent
Evaluated models
GPT-5.5, GPT-5.4 mini, Claude-Sonnet-4.6 and others7
- AttacksPrompt injection
- TechniqueDormant triggers
- ComponentMemory
Paper analysisRead analysis
AttacksarXiv 2608.25776Aug 26
EvoMal attack uses self-poisoning of self-evolving coding agents to achieve worm-like propagation
提出 EVOMAL，诱导自演化编码 Agent 在创建技能时自我投毒并蠕虫传播
arXiv
2608.25776
Published
Aug 26
Layer
Application layer
Setting
coding agent
Evaluated models
Devstral-Small-2, Gemma-4-31B-IT, Qwen3-Coder-Next and others7
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentMCP and skills
Abstract论文指出了自演化智能体模仿检索代码的自我毒化漏洞，提出 EVOMAL 攻击与防御，阐明了供应链新风险。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2608.12851Aug 13
The researchers propose SkillMisevo-Bench to quantify the risk of skill mis-evolution in self-improving agents.
提出 SkillMisevo-Bench，量化自改进 Agent 的技能误演化
arXiv
2608.12851
Published
Aug 13
Layer
Application layer
Setting
coding agent
Evaluated models
Claude Code, Codex, Hermes and others4
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentMemory
Paper analysisRead analysis
AttacksarXiv 2608.09577Aug 10
ElasticBack：通过触发词与规则耦合在 LLM Agent 技能中植入条件后门
提出 ElasticBack，在单个 Agent 技能中植入条件后门
arXiv
2608.09577
Published
Aug 10
Layer
Application layer
Setting
coding agent
Evaluated models
GLM-5.2, MiniMax-M3, GPT-5.4 and others4
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentMCP and skills
AbstractElasticBack 用冻结规则加遗传搜索演化 trigger，在单 skill 上做条件后门。
ElasticBack 是放在 LLM agent 单个 skill 里的条件后门攻击。。
Paper analysisRead analysis
AttacksarXiv 2608.06862Aug 7
SynChain：诱导计算机使用 Agent 自建攻击链并跨任务持久化
提出 SYNCHAIN，诱导计算机使用 Agent 自建跨任务持久攻击链
arXiv
2608.06862
Published
Aug 7
Layer
Application layer
Setting
Browser and computer use, coding agent
Evaluated models
Qwen3.5-9B, Llama-3.1-8B, Ministral-3-8B
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentMCP and skills
Abstract论文揭示了 CUA 自合成工件带来的内部供应链威胁，验证了潜伏载荷的跨任务传播能力与防御局限。
Paper analysisRead analysis
AttacksarXiv 2607.14493Jul 16
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
Jul 16
Layer
Application layer
Setting
LLM applications
Evaluated models
GPT-4o, Claude 3.5 Sonnet, Llama-3-70B-Instruct
- AttacksPrompt injection
- TechniqueEncoding and obfuscation
- ComponentGuardrails and judges
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
AttacksarXiv 2606.09700Jun 9
The researchers propose the HPAA attack: bypassing 13 content moderation systems with typographic visual cues
提出 HPAA 攻击，用排版视觉线索绕过内容审核系统
arXiv
2606.09700
Published
Jun 9
Layer
Prompt layer
Setting
LLM applications
Attacker
Black-box
Evaluated models
Llama-Guard-3-8B, Perspective API, ShieldGemma-2B and others13
- AttacksDetection evasion
- TechniqueEncoding and obfuscation
- ComponentGuardrails and judges
Abstract揭示了排版视觉线索与分词差异带来的审核盲区，HPAA 以极低查询实现近乎完全规避。
这篇论文研究了人类视觉感知与大模型分词处理机制之间的差异所导致的内容审核漏洞。
Paper analysisRead analysis
AttacksarXiv 2606.09499Jun 8
研究者提出针对世界模型的机器人学习管线投毒攻击
提出 VPH 与 VTH，只改视频帧劫持世界模型并投毒下游策略
arXiv
2606.09499
Published
Jun 8
Layer
Training and data layer
Setting
Embodied systems and robotics
Evaluated models
Cosmos-Predict 2.5, Cosmos-Predict 2.5 AC, Dino WM and others4
- AttacksPoisoning and backdoors
- TechniqueGradient and representation optimization
- ComponentVideo
Abstract只改视频帧经世界模型投毒。
只改看似安全的视频帧，经世界模型之后才改变下游机器人策略。 问题在于世界模型进入数据生成后，数据集检查看到的仍是安全演示，合成轨迹却可以变危险。
Paper analysisRead analysis
AttacksarXiv 2605.24421May 23
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
May 23
Layer
Application layer
Setting
LLM applications
Evaluated models
gpt-4o-mini
- AttacksPrompt injection
- TechniqueConversation structure manipulation
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
AttacksarXiv 2605.22258May 21
CITA：面向中文隐式毒性的攻击与防御数据生成框架
提出 CITA 三阶段框架，生成可绕过毒性检测器的中文隐式毒性文本
arXiv
2605.22258
Published
May 21
Layer
Application layer
Setting
LLM applications
Evaluated models
Tencent, Baidu, Gemini 3.1 Pro and others11
- AttacksDetection evasion
- TechniqueEncoding and obfuscation
- ComponentGuardrails and judges
Paper analysisRead analysis
AttacksarXiv 2604.03890Apr 5
From prompts to physical actions: structured backdoor attacks against LLM-mediated robot control systems
Propose a structured JSON backdoor that triggers malicious actions in robot control via a poisoned LoRA
arXiv
2604.03890
Published
Apr 5
Layer
Training and data layer
Setting
Embodied systems and robotics, LLM applications
Evaluated models
Llama-3.1-8B-Instruct, Gemma-2-9B-IT, DeepSeek-R1-Distill-Llama-8B and others4
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentFine-tuning and open weights
Abstract推理层后门传不到执行。
Paper analysisRead analysis

AI Safety Daily

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