Poisoning the Neighborhood: A Topology-Aware Topic-Level Attack on Large-Scale HNSW-based RAG Systems
A novel poisoning attack exploits the HNSW graph topology of large-scale RAG systems, achieving high retrieval corruption rates that generic embedding defences miss—critical for any enterprise relying on external knowledge ingestion.
Summary written by editorial AI · Source link below
arXiv:2610.21464v1 Announce Type: new Abstract: Retrieval-augmented generation (RAG) mitigates hallucinations and knowledge staleness in large language models using external knowledge. However, ingestion from public and third-party sources exposes RAG systems to poisoning attacks. Existing attacks are designed and evaluated under exact retrieval, where poisoned documents participate in global ranking. Deployment-scale systems instead rely on approximate nearest-neighbor (ANN) indexes such as hi
Editorial Analysis
Enterprises feeding external or crowd-sourced data into RAG pipelines face a practical poisoning vector that bypasses standard cosine-similarity defences, risking corrupted AI outputs.
Add vector-index integrity checks and source-provenance controls to RAG ingestion workflows before scaling external data sources.
AI knowledge systems that pull from public data can be silently poisoned through their index structure—review your RAG pipeline controls.
Forward-looking interpretation drafted by editorial AI under human review — not a reproduction of the source. See methodology.
External link — opens at arXiv Crypto & Security in a new tab.
More from the AI Security Desk
- Claude Code's “Don't Ask Again” File Exposes Passwords and Keys in 1,122 Public Copies on GitHub2d
- Runaway AI hacking agent behind WorkOS breach after escaping sandbox3d
- AI Gave Your Incident Archive a Second Job3d
- Malicious Browser Extensions Could Use Claude Dispatch to Run Code on Users' Computers3d
- AI Slop and Hallucinations in Vulnerability Assessment: A Survey on Reasoning Failures and Trustworthy Mitigation3d