Metraipolis › archive › dc_337d8081e37f
Intent-Hiding Jailbreaks: An Information-Theoretic Framework for Compositional Attacks — arXiv:2610.02302
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.02302
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-05T23:06:18.032Z
- SHA-256 of the archived text
e81cffb80c6ea8fd36a0880f7ab5bf0ac8f86a9876c17ae1a95d86a7c56987c7- Stable URL of this page
https://metraipolis.com/documento/dc_337d8081e37f
How to cite
Intent-Hiding Jailbreaks: An Information-Theoretic Framework for Compositional Attacks — arXiv:2610.02302 — arXiv, https://arxiv.org/abs/2610.02302 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_337d8081e37f (Metraipolis archive dc_337d8081e37f, sha256 e81cffb80c6ea8fd36a0880f7ab5bf0ac8f86a9876c17ae1a95d86a7c56987c7)
BibTeX
@misc{dc337d8081e37f,
title = {Intent-Hiding Jailbreaks: An Information-Theoretic Framework for Compositional Attacks — arXiv:2610.02302},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.02302},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_337d8081e37f; sha256 e81cffb80c6ea8fd36a0880f7ab5bf0ac8f86a9876c17ae1a95d86a7c56987c7},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.02302. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Recent work has shown that large language models (LLMs) can be vulnerable to jailbreak attacks in which harmful intent is obscured through composition with benign tasks. A harmful request refused in isolation may elicit a different response when embedded within a larger, seemingly benign query. We study these compositional intent-hiding jailbreaks from an information-theoretic perspective. Our formulation associates each task with an estimated probability of being judged harmful: the average over the full task collection defines the prior probability of harmful intent, while the average over a selected bundle containing the target defines the posterior. Selecting auxiliary tasks so that these averages agree, which we call prior-posterior matching, leaves the estimated intent unchanged even though the harmful target remains in the bundle. We study two settings that differ in whether query construction is part of the optimization. In the query-independent setting, tasks are selected without regard to how they will be expressed in the final query. We show that exact prior-posterior matching under a bundle-size constraint is computationally hard, derive an optimal water-filling solution for fractional weights, and characterize the smallest bundle satisfying a prescribed safety threshold. In the query-dependent setting, task selection and query construction are considered jointly, and intent concealment and target preservation are evaluated on the resulting query. We evaluate jailbreak effectiveness and preservation of the target behavior across bundle sizes, query generators, and several open-source models. These results show that compositional queries can elicit target behaviors beyond the direct-request baseline under the evaluated search budgets, while revealing a trade-off: as bundle size increases, response-level target preservation tends to decrease for several models. Tema degli esploratori: fiducia digitale e crittografia (elenco arXiv cs.CR, nuovi lavori del giorno). Categorie arXiv: Cryptography and Security (cs.CR) ; Information Theory (cs.IT); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.02302 — 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 2026-10-05T23:01:08.528Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).