Lune

EMNLP Findings2025

RevPRAG: Revealing Poisoning Attacks in Retrieval-Augmented Generation through LLM Activation Analysis

Xue Tan, Hao Luan, Mingyu Luo, Xiaoyan Sun, Ping Chen, Jun Dai

摘要

Retrieval-Augmented Generation (RAG) enriches the input to LLMs by retrieving information from the relevant knowledge database, enabling them to produce responses that are more accurate and contextually appropriate.It is worth noting that the knowledge database, being sourced from publicly available channels such as Wikipedia, inevitably introduces a new attack surface.RAG poisoning attack involves injecting malicious texts into the knowledge database, ultimately leading to the generation of the attacker's target response (also called poisoned response).However, there are currently limited methods available for detecting such poisoning attacks.We aim to bridge the gap in this work by introducing RevPRAG, a flexible and automated detection pipeline that leverages the activations of LLMs for poisoned response detection.Our investigation uncovers distinct patterns in LLMs' activations when generating poisoned responses versus correct responses.Our results on multiple benchmarks and RAG architectures show our approach can achieve a 98% true positive rate, while maintaining a false positive rate close to 1%.Query Answer LLM Prompt Context: [...] Everest stands the tallest [...

问 Lune

几分钟读透这篇 Paper,不用再花一下午。

Lune 能读完全文,也能读引用它的顶会 Paper,讲清它的结论、证据和局限,每一点都注明出自哪篇 Paper。不用硬啃,也不靠猜。

已有 5 篇顶会 Paper 引用了它。别等审稿人问起,先弄清它们发现了什么。

可以从这些问题问起

Lune 可以查阅

全文 · 引用它的顶会 Paper · 相关的 Paper

获取答案

免费开始,无需绑卡

脉络

  • InceptionRAG: Stealthy Poisoning Attack Against Retrieval-Augmented Generation 引用了 Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval
  • InceptionRAG: Stealthy Poisoning Attack Against Retrieval-Augmented Generation 引用了 Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection
  • Making Retrieval-Augmented Language Models Robust to Irrelevant Context 引用了 Measuring and Narrowing the Compositionality Gap in Language Models
它引用的引用它的从左到右按年份排列,引用越多,位置越高。

引用它的顶会 Paper5

  1. SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAGXiaonan Si, Meilin Zhu, Simeng Qin, Lijia Yu 等NeurIPS 2025 · 16 次引用
  2. Interpretation Meets Safety: A Survey on Interpretation Methods and Tools for Improving LLM SafetySeongmin Lee, Aeree Cho, Grace C. Kim, Shengyun Peng 等EMNLP 2025 · 1 次引用
  3. Mitigating GenAI-Powered Evidence Pollution for Out-Of-Context Misinformation DetectionZehong Yan, Peng Qi, Wynne Hsu, Mong-Li LeeICDE 2026 · 1 次引用
  4. InceptionRAG: Stealthy Poisoning Attack Against Retrieval-Augmented GenerationJiachang Zhang, Min Chen, Xiao Ren, Zhenyong Zhang 等CCS 2026
  5. One Single Hub Text Breaks CLIP: Identifying Vulnerabilities in Cross-Modal Encoders via HubnessHiroyuki Deguchi, Katsuki Chousa, Yusuke SakaiACL 2026

它引用的 Paper15

  1. Self-RAG: Learning to Retrieve, Generate, and Critique through Self-ReflectionAkari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil 等ICLR 2024 · 1,798 次引用
  2. Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text RetrievalLee Xiong, Chenyan Xiong, Ye Li, Kwok-Fung Tang 等ICLR 2021 · 1,547 次引用
  3. On Adaptive Attacks to Adversarial Example DefensesFlorian Tramèr, Nicholas Carlini, Wieland Brendel, Aleksander MadryNeurIPS 2020 · 1,026 次引用
  4. Making Retrieval-Augmented Language Models Robust to Irrelevant ContextOri Yoran, Tomer Wolfson, Ori Ram, Jonathan BerantICLR 2024 · 361 次引用
  5. Measuring and Narrowing the Compositionality Gap in Language ModelsOfir Press, Muru Zhang, Sewon Min, Ludwig Schmidt 等EMNLP Findings 2023 · 296 次引用

相关 Paper

  1. On the Vulnerability of Applying Retrieval-Augmented Generation within Knowledge-Intensive Application DomainsXun Xian, Ganghua Wang, Xuan Bi, Rui Zhang 等ICML 2025
  2. RIPRAG: Hack a Black-box Retrieval-Augmented Generation Question-Answering System with Reinforcement LearningMeng Xi, Sihan Lv, Yechen Jin, Guanjie Cheng 等ACL Findings 2026
  3. KEPo: Knowledge Evolution Poison on Graph-based Retrieval-Augmented GenerationQizhi Chen, Chao Qi, Yihong Huang, Muquan Li 等WWW 2026
  4. PR-Attack: Coordinated Prompt-RAG Attacks on Retrieval-Augmented Generation in Large Language Models via Bilevel OptimizationYang Jiao, Xiaodong Wang, Kai YangSIGIR 2025 · 6 次引用
  5. Traceback of Poisoning Attacks to Retrieval-Augmented GenerationBaolei Zhang, Haoran Xin, Minghong Fang, Zhuqing Liu 等WWW 2025 · 20 次引用
黄昏的海面,两侧是细线勾勒的悬崖