Metraipolis › archive › dc_23b7d4e7883d

Stealthy Poisoning Attacks via Reused KV Cache of Reasoning — arXiv:2610.22873

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.22873
License (as archived)
CC0-1.0
Archived
2026-10-19T08:46:38.013Z
SHA-256 of the archived text
a61282657f26ff490c33224865543450d473417a57fb07a78aefd32a477ef310
Stable URL of this page
https://metraipolis.com/documento/dc_23b7d4e7883d

How to cite

Stealthy Poisoning Attacks via Reused KV Cache of Reasoning — arXiv:2610.22873 — arXiv (en), https://arxiv.org/abs/2610.22873 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_23b7d4e7883d (Metraipolis archive dc_23b7d4e7883d, sha256 a61282657f26ff490c33224865543450d473417a57fb07a78aefd32a477ef310)
BibTeX
@misc{dc23b7d4e7883d,
  title = {Stealthy Poisoning Attacks via Reused KV Cache of Reasoning — arXiv:2610.22873},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.22873},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_23b7d4e7883d; sha256 a61282657f26ff490c33224865543450d473417a57fb07a78aefd32a477ef310},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22873. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Modern large language model (LLM) systems reuse key-value (KV) caches to accelerate inference. Recent work has shown that reusing KV caches from reasoning traces can also improve task performance, similar to retrieval-augmented generation (RAG). We demonstrate that this form of KV cache reuse introduces new poisoning vulnerabilities. We propose R$^2$ECAP (Reused Reasoning-Cache Poisoning), a black-box attack that plants poisoned reasoning traces solely through interactions with a system's chat interface. R$^2$ECAP constructs two traces: a preceding trace that steers the model toward an adversarial target and a retained trace that carries this target. Because the retained trace is generated conditioned on the preceding trace, its KV cache preserves residual adversarial influence, even after the preceding trace and its cache entries are discarded. We evaluate R$^2$ECAP on three open-weight LLMs and three datasets. R$^2$ECAP achieves up to 79% attack success rate (ASR) and outperforms HijackKV, a state-of-the-art white-box KV poisoning attack, in 8 of 9 resource-constrained settings. It also achieves 46% ASR in a realistic reasoning-trace retrieval system. Finally, R$^2$ECAP remains stealthy, with only 1% of its retained responses detected by a popular prompt injection detector.

Tema degli esploratori: fiducia digitale e crittografia (elenco arXiv cs.CR, nuovi lavori del giorno).
Categorie arXiv: Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2610.22873 — 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 2026-10-19T08:41:54.767Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).