Metraipolis › archive › dc_87aa64c13cec
ToxicRAG: Compromising Retrieval-Augmented Generation Systems via Single-Shot Knowledge Poisoning Attacks — arXiv:2609.11082
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.11082
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T05:46:27.280Z
- SHA-256 of the archived text
ca471c080deaf19f7123a188bdfdc175457253f4c9684b35c09793c6708dbf6e- Stable URL of this page
https://metraipolis.com/documento/dc_87aa64c13cec
How to cite
ToxicRAG: Compromising Retrieval-Augmented Generation Systems via Single-Shot Knowledge Poisoning Attacks — arXiv:2609.11082 — arXiv (en), https://arxiv.org/abs/2609.11082 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_87aa64c13cec (Metraipolis archive dc_87aa64c13cec, sha256 ca471c080deaf19f7123a188bdfdc175457253f4c9684b35c09793c6708dbf6e)
BibTeX
@misc{dc87aa64c13cec,
title = {ToxicRAG: Compromising Retrieval-Augmented Generation Systems via Single-Shot Knowledge Poisoning Attacks — arXiv:2609.11082},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.11082},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_87aa64c13cec; sha256 ca471c080deaf19f7123a188bdfdc175457253f4c9684b35c09793c6708dbf6e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.11082. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Retrieval-Augmented Generation (RAG) can ground large language model (LLM) outputs in external evidence, but it also exposes the system to knowledge poisoning. Representative attacks use multiple injected documents or templates that directly assert a target answer. We present ToxicRAG, a one-document-per-target attack that expresses misinformation as a coherent knowledge-update narrative. The generated document first acknowledges the previously accepted answer, introduces fabricated events that appear to invalidate it, and then attributes the attacker-selected answer to a set of purported authorities. An answer-focused self-validation loop optionally revises a candidate when a surrogate language model does not reproduce the target answer. We evaluate the attack on 100 target questions from each of Natural Questions, HotpotQA, and MS-MARCO, using four victim LLMs and four dense retrievers. In the sampled-corpus setting reported in this paper, ToxicRAG obtains ASRs between 0.61 and 0.91 across the twelve dataset--model combinations. It matches or exceeds the strongest evaluated baseline in every combination, with margins ranging from 0 to 11 percentage points. These results show that narrative-form poisoned documents can remain influential under the evaluated RAG configurations and motivate further study of factual consistency and source provenance in RAG systems. Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-11 in cs.CR). Categorie arXiv: Cryptography and Security (cs.CR). Pagina del lavoro: https://arxiv.org/abs/2609.11082 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-16T05:30:39.384Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-11?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).