Metraipolis › archive › dc_675aeeed2c4a

SEAL: Sealed Queries for Local Sanitization against Indirect Prompt Injection — arXiv:2610.39758

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.39758
License (as archived)
CC0-1.0
Archived
2026-11-01T02:21:19.054Z
SHA-256 of the archived text
0f0cf50d156a1772e5309832d27fd21684075c1c7b0be8a1edd4226bbac00e56
Stable URL of this page
https://metraipolis.com/documento/dc_675aeeed2c4a

How to cite

SEAL: Sealed Queries for Local Sanitization against Indirect Prompt Injection — arXiv:2610.39758 — arXiv (en), https://arxiv.org/abs/2610.39758 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_675aeeed2c4a (Metraipolis archive dc_675aeeed2c4a, sha256 0f0cf50d156a1772e5309832d27fd21684075c1c7b0be8a1edd4226bbac00e56)
BibTeX
@misc{dc675aeeed2c4a,
  title = {SEAL: Sealed Queries for Local Sanitization against Indirect Prompt Injection — arXiv:2610.39758},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.39758},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_675aeeed2c4a; sha256 0f0cf50d156a1772e5309832d27fd21684075c1c7b0be8a1edd4226bbac00e56},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.39758. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Indirect prompt injection threatens local LLM agents through untrusted tool outputs, motivating defenses that suppress attacks while preserving task evidence under limited inference budgets. We introduce SEAL, which learns a soft prompt and scoring readout offline to sanitize tool outputs using a frozen task model. SEAL constructs scoring queries from trusted initial context and the learned prompt, keeping the reference independent of inspected tool text. It scores tokens without normalization across context tokens and uses preservation regularization to encourage retention of useful evidence during localized deletion. This design requires neither defense-generated tokens nor a separate verifier model. Across AgentDojo and AgentDyn, SEAL suppresses attacks while improving task completion under attack. Specifically, on AgentDojo, SEAL reduces attack success from 34.6\% to 0.1\% and increases Joint---task completion without attack success---from 51.4\% to 77.5\%, while using approximately $127\times$ fewer defense-related tokens than PISanitizer, a state-of-the-art prompt injection defense. Replication on a second backbone with the same training recipe supports applicability across model families.

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/2610.39758 — 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 2026-11-01T02:12:13.399Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).