AI Security

CamoDocs: A Poisoning Attack Against Retrieval-Augmented Language Models Using Camouflaged Documents

CamoDocs demonstrates a stealth poisoning technique against RAG systems where camouflaged malicious documents evade current defences — a growing concern as enterprises connect LLMs to external knowledge bases.

Summary written by editorial AI · Source link below

Filed by arXiv Crypto & Security31 August 20261 min readRead at source ↗

arXiv:2608.28389v1 Announce Type: new Abstract: Retrieval-augmented generation (RAG) augments LLMs with external documents, but public or user-editable sources expose RAG systems to data poisoning: attackers can inject malicious documents to steer outputs toward targeted answers. Existing poisoning attacks often rely on query inclusion, inserting the target query into poisoned documents to improve retrieval; however, this creates lexical and embedding-space artifacts that make them easy to filt

Editorial Analysis

Why it matters

Enterprises connecting LLMs to external or collaboratively edited knowledge bases face a real risk that poisoned documents can steer model outputs toward attacker-chosen answers without detection.

What to do

Implement document-integrity verification and retrieval-anomaly monitoring for all external data sources feeding RAG-based enterprise applications.

Board brief

RAG-augmented AI systems that pull from external document stores can be silently poisoned to produce attacker-controlled outputs, requiring new data-provenance controls.

Forward-looking interpretation drafted by editorial AI under human review — not a reproduction of the source. See methodology.

Continue at the source

Read the full report at arXiv Crypto & Security

External link — opens at arXiv Crypto & Security in a new tab.

Continue with

Editorial Analysis

More from the AI Security Desk