Metraipolis › archive › dc_231d7969a49b

Detecting Unseen Jailbreak Sources: A Multi-Source Conformal Detection Perspective — arXiv:2610.09167

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.09167
License (as archived)
CC0-1.0
Archived
2026-10-08T12:40:16.669Z
SHA-256 of the archived text
129edd925848c47d3c9808d6da5134fce9f63293962bd99e75f823392fa52499
Stable URL of this page
https://metraipolis.com/documento/dc_231d7969a49b

How to cite

Detecting Unseen Jailbreak Sources: A Multi-Source Conformal Detection Perspective — arXiv:2610.09167 — arXiv (en), https://arxiv.org/abs/2610.09167 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_231d7969a49b (Metraipolis archive dc_231d7969a49b, sha256 129edd925848c47d3c9808d6da5134fce9f63293962bd99e75f823392fa52499)
BibTeX
@misc{dc231d7969a49b,
  title = {Detecting Unseen Jailbreak Sources: A Multi-Source Conformal Detection Perspective — arXiv:2610.09167},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.09167},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_231d7969a49b; sha256 129edd925848c47d3c9808d6da5134fce9f63293962bd99e75f823392fa52499},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09167. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Jailbreak defenses for large language models are usually calibrated on a fixed set of known attack methods. However, new attack methods keep appearing after deployment, and existing detectors are observed to degrade under the new attacks (Piet et al., 2025). This paper studies how to detect, from a stream of attack prompts, that the prompts come from an attack source outside the known ones. We formulate it as a sequential test whose null hypothesis is that the stream comes from one of the known sources. Methodology-wise, we compute a conformal p-value against each known source, and reject the null only when the evidence against every known source is large. Using these source-wise p-values, we build Persist-3 and analyze its stationary false-alarm probability and detection power. The technical contribution of Persist-3 lies in a detection guarantee that uses only a finite window of recent observations. Our theoretical analysis reveals that its detection probability approaches one exponentially fast as the window grows, and thus a short window suffices to detect unseen sources that are well separated from all known sources. Empirically, we validate Persist-3 on top of two recent jailbreak defenses, MTK (Zhang et al., 2026) and Prompt Guard 2 (Chennabasappa et al., 2025), and observe a significant defense enhancement.

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv stat.AP, nuovi lavori del giorno).
Categorie arXiv: Applications (stat.AP).
Pagina del lavoro: https://arxiv.org/abs/2610.09167 — 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 2026-10-08T12:15:30.214Z dalla pagina https://arxiv.org/list/stat.AP/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).