Metraipolis › archive › dc_e6829fd2dd9a
Your Model Is Leaking: Covert Information Transfer through LLM Residual Streams — arXiv:2609.27996
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.27996
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T10:13:25.406Z
- SHA-256 of the archived text
8a13b2c9b5ee9455dc174496e46a46016b7b077674b138e3cc58763eb92c1061- Stable URL of this page
https://metraipolis.com/documento/dc_e6829fd2dd9a
How to cite
Your Model Is Leaking: Covert Information Transfer through LLM Residual Streams — arXiv:2609.27996 — arXiv (en), https://arxiv.org/abs/2609.27996 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e6829fd2dd9a (Metraipolis archive dc_e6829fd2dd9a, sha256 8a13b2c9b5ee9455dc174496e46a46016b7b077674b138e3cc58763eb92c1061)
BibTeX
@misc{dce6829fd2dd9a,
title = {Your Model Is Leaking: Covert Information Transfer through LLM Residual Streams — arXiv:2609.27996},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.27996},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e6829fd2dd9a; sha256 8a13b2c9b5ee9455dc174496e46a46016b7b077674b138e3cc58763eb92c1061},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.27996. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Privacy-sensitive organizations may run large language models (LLMs) in restricted or air-gapped environments while exporting selected diagnostic artifacts. We show that a compromised runtime component can hide sensitive information in intermediate activations that are allowed to leave the restricted environment. An offline observer can recover this information with a simple linear decoder. The attack requires no model retraining or weight modification, no attacker-controlled egress, and no control over the recorder or transfer process. We introduce a residual-stream covert-channel attack that maps messages to codewords and injects them into an intermediate residual stream through a compromised runtime hook. To maintain recoverability, the injection strength is scaled with the local residual norm using the signal-to-residual-norm ratio. Across eleven models from seven architecture families, our evaluation shows 91--100% recovery on nine models with KL divergence 0.001--0.007, while evaluated activation-level detectors remain close to random guessing (AUC <= 0.56). Tested post-hoc defenses do not reliably eliminate the channel. Thus, an activation artifact can be schema-valid while carrying information that is not authorized to cross the boundary. Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-24 in cs.CR). Categorie arXiv: Cryptography and Security (cs.CR). Pagina del lavoro: https://arxiv.org/abs/2609.27996 — 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 2026-10-08T09:26:35.720Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).