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

Oct 12 · Monday

All papers

Found 10 papers

Filter5

Attack type

Poisoning and backdoors10 Model tampering7 Extraction and theft6 Detection evasion3

Attack technique

Dormant triggers4

Impact

Output manipulation8

System type

Models and APIs3 LLM applications4 AI Agent3

Attack surface / entry point

User input3 Retrieval corpus4 Weights and gradients5 Distribution channels4

Layer

Model layer3 Application layer7 Training and data layer24

System component

MCP and skills41 Memory43 RAG22 Guardrails and judges8 Monitors6 Reasoning chains3 Permissions and sandboxing11 Fine-tuning and open weights6 Embeddings4 Multimodal3

Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 233papers have no tags yet and are not included in filtered results.

Attack type：Poisoning and backdoors Layer：Model layer Layer：Application layer System component：Fine-tuning and open weights System component：Embeddings Clear all

Another 233papers 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
DefensesarXiv 2609.39548Sep 30
NDDL：用正常扩散动力学为文生图模型做后门防御
提出 NDDL，用良性扩散轨迹检测文生图后门提示并定位触发 token
arXiv
2609.39548
Published
Sep 30
Layer
Model layer
Setting
Models and APIs
Evaluated models
Stable Diffusion v1.5, Stable Diffusion XL, Stable Diffusion v3.5 and others4
- DefensesMonitoring and auditing
- AttacksPoisoning and backdoors
- ComponentImage generation
AbstractNDDL 从良性扩散轨迹学习正常转移，用动态不一致检测后门并定位触发 token。
NDDL 是面向文生图扩散模型的后门防御：防御者只有白盒模型访问和良性提示词，没有后门先验，需要区分后门提示词并找出触发 token。
Paper analysisRead analysis
DefensesarXiv 2610.00348Sep 30
NEEDLE：通过权重正交化移除 LLM 后门
提出 NEEDLE，用权重正交化在已知触发器时去除大模型后门
arXiv
2610.00348
Published
Sep 30
Layer
Model layer
Setting
Models and APIs
Evaluated models
Gemma-3-4B-IT, Qwen3-4B-Instruct-2507, Gemma-3-12B-IT
- DefensesModel hardening
- AttacksPoisoning and backdoors
- ComponentFine-tuning and open weights
AbstractNEEDLE 用闭式权重编辑去掉已知触发器的后门，并保住拒绝子空间。
Paper analysisRead analysis
AttacksarXiv 2609.37468Sep 26
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
Sep 26
Layer
Application layer
Setting
AI Agent
Evaluated models
E5, Contriever-MSMARCO
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentRAG
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
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
DefensesarXiv 2607.04379Jul 6
PA-LLM-RAG 框架的知识库投毒攻击与防御
提出策略 RAG 的知识库投毒攻击与检测防御 CLD-KB
arXiv
2607.04379
Published
Jul 6
Layer
Application layer
Setting
LLM applications
Evaluated models
Llama 3.2
- DefensesDetection and filtering
- AttacksPoisoning and backdoors
- ComponentRAG
Paper analysisRead analysis
AttacksarXiv 2605.28074May 27
SilentRetrieval: Hijacking retrieval-augmented generation with semantically preserved adversarial data poisoning
提出 SilentRetrieval，以语义保持对抗文档投毒劫持 RAG 检索与生成
arXiv
2605.28074
Published
May 27
Layer
Application layer
Setting
LLM applications
Evaluated models
Llama-2-7B-Chat, Mistral-7B-Instruct-v0.2, Qwen-7B-Chat and others10
- AttacksPoisoning and backdoors
- TechniqueGradient and representation optimization
- ComponentRAG
Abstract两阶段投毒在单文档协议下兼顾检索劫持、生成操纵与近良性困惑度。
SilentRetrieval 是针对 RAG 的两阶段语料投毒，在每查询一篇、合成目标答案的统一协议下，同时处理检索劫持、生成操纵和流畅度。
Paper analysisRead analysis
DefensesarXiv 2602.03085Feb 3
Triggers in the Haystack: Extracting and Reconstructing LLM Backdoor Triggers
提出推理扫描器，在仅有模型文件时检测后门并重构触发词
arXiv
2602.03085
Published
Feb 3
Layer
Model layer
Setting
Models and APIs
Evaluated models
gemma-3-270m-it, DeepSeek-R1-Distill-Qwen-1.5B, Phi-4-mini-instruct and others8
- DefensesMonitoring and auditing
- AttacksPoisoning and backdoors
- ComponentFine-tuning and open weights
Abstract用记忆泄漏和内部信号重构触发词。
这是一个只做推理的 sleeper agent 扫描器：在不知触发词和目标行为时，检测因果语言模型里的条件后门并重构触发词。。问题在于，后门只在触发词出现时偏离基线，而开放仓库和网页投毒使检测方通常没有触发词、目标或激活语料。
Paper analysisRead analysis
AttacksarXiv 2602.00364Jan 31
"Someone hid it": Query-agnostic black-box attacks against LLM retrieval
提出查询无关黑盒攻击 DQ-A，用可迁移文末后缀降低文档检索排名
arXiv
2602.00364
Published
Jan 31
Layer
Application layer
Setting
LLM applications
Evaluated models
Qwen-1.5-7B-Instruct, SFR-Embedding-Mistral, E5-Mistral-7B-Instruct and others10
- AttacksPoisoning and backdoors
- TechniqueGradient and representation optimization
- ComponentRAG
AbstractDQ-A 以无查询黑盒后缀压低 LLM 检索排名，并在 BRIGHT 上比较三种基线。
DQ-A 是一种 query-agnostic、complete black-box 的文档隐藏攻击，对象是用最后隐层相似度匹配的 LLM 检索。
Paper analysisRead analysis
AttacksarXiv 2512.01335Dec 1, 2025
EmoRAG 研究揭示 RAG 检索对表情符号扰动的脆弱性
提出 EmoRAG，向查询与知识库注入同一表情以劫持 RAG 检索
arXiv
2512.01335
Published
Dec 1, 2025
Layer
Application layer
Setting
LLM applications
Evaluated models
SPECTER, Contriever, Qwen2-7B and others11
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentRAG
AbstractEmoRAG 用表情联动查询与库文本。
EmoRAG 测量 RAG 对符号扰动的反应：同一表情同时出现在查询和库文本里时，检索会偏离语义匹配。作者要检查的是，检索是否真由语义相关性决定。
Paper analysisRead analysis

You have reached the end

AI Safety Daily

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