Newest papers

RAGLeak: Supply-Chain Privacy Backdoors in Retrieval-Augmented Generation

Yuchen Chen, Wei Cheng, Yuan Xiao, Junda He, Jieke Shi, 欧阳萍萍, Quanjun Zhang, Chunrong Fang, David Lo, Zhenyu Chen · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.25937 · cited by 0 works (OpenAlex) · open-access PDF

German

Retrieval-Augmented Generation (RAG) ermöglicht großen Sprachmodellen den Zugang zu externen Kenntnissen während der Schlussfolgerung, aber seine Abhängigkeit von externen Wissensbasen führt auch neue Angriffsoberflächen ein. Bestehende Daten-Extraktionsangriffe gegen RAG-Systeme verlassen sich entweder auf eine prompt-Injektion, die stark von der Anweisungsnachfolgungsfähigkeit des Generators abhängt, oder auf Hintertüren, die abgerufene Dokumente reproduzieren oder paraphrasieren, aber kämpfen, um Extraktionsfidelität und Stealthiness auszugleichen. In diesem Papier schlagen wir RAGLeak vor, eine Lieferketten-Privatsphäre-Backdoor-Angriff, die die Informationslücke von der direkten Text

Machine translation (Meta M2M100); the English original follows.

Abstract

Retrieval-Augmented Generation (RAG) enables large language models to access external knowledge during inference, but its reliance on external knowledge bases also introduces new attack surfaces. Existing data extraction attacks against RAG systems rely on either prompt injection, which heavily depends on the generator's instruction-following capability, or backdoors that reproduce or paraphrase retrieved documents but struggle to balance extraction fidelity and stealthiness. In this paper, we propose RAGLeak, a supply-chain privacy backdoor attack that decouples information leakage from direct textual exposure. When triggered, the generator reversibly obfuscates retrieved private documents for high-fidelity recovery by an auxiliary LLM. Across three model families and three application domains, RAGLeak achieves up to $100\%$ ASR with limited impact on benign utility and at most a $10$-percentage-point ASR drop under representative detection-based defenses. The backdoor remains effective after clean fine-tuning and across model scales and versions, revealing an underexplored supply-chain privacy risk in RAG systems. The source code is available at https://anonymous.4open.science/r/RAGLeak-4B67.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Backdoor, Computer science, Computer security, Paraphrase, Generator (circuit theory), Source code

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange