Metraipolis › archive › dc_7e4b5ba3735f

SIR: Self-improving Red-teaming for Compute Use Agents — arXiv:2608.30207

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.30207
License (as archived)
CC0-1.0
Archived
2026-10-17T07:52:28.713Z
SHA-256 of the archived text
7ea3a5e9ab077696e7ac6a8ba4637148d8d3b06e3e07a76bf98e5585362f191f
Stable URL of this page
https://metraipolis.com/documento/dc_7e4b5ba3735f

How to cite

SIR: Self-improving Red-teaming for Compute Use Agents — arXiv:2608.30207 — arXiv (en), https://arxiv.org/abs/2608.30207 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7e4b5ba3735f (Metraipolis archive dc_7e4b5ba3735f, sha256 7ea3a5e9ab077696e7ac6a8ba4637148d8d3b06e3e07a76bf98e5585362f191f)
BibTeX
@misc{dc7e4b5ba3735f,
  title = {SIR: Self-improving Red-teaming for Compute Use Agents — arXiv:2608.30207},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.30207},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7e4b5ba3735f; sha256 7ea3a5e9ab077696e7ac6a8ba4637148d8d3b06e3e07a76bf98e5585362f191f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.30207. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Computer-use agents (CUAs) are agents powered by vision-language models (VLMs) that perceive a screen and operate an operating system through mouse, keyboard, and terminal interactions to automate everyday digital tasks. Their exposure to untrusted content creates a risk of indirect prompt injection (IPI), where an adversary embeds instructions in content the agent reads to redirect it toward actions that violate the user's intent. Evaluations based on fixed, hand-written injections may underestimate the risk posed by adaptive adversaries. We present SIR, a black-box self-improving IPI framework that (i) composes task-specific injections from a small library of reusable red-teaming principles stated in plain language and (ii) uses an iterative feedback loop to analyze unsuccessful attack trajectories and distill new, named principles that are retained in a shared library and reused across tasks. We target operating-system-level compromise and evaluate outcomes through deterministic checks on filesystem, service, and permission state rather than an LLM judge. An attack counts as successful only when both the adversarial objective and the benign user task are completed in the same execution. We evaluate three frontier CUAs, allowing SIR up to 10 attack attempts per case. It achieves joint attack success rates of 24% on Claude Opus 4.8 and 28% on Gemini 3.5 Flash, compared with 4% and 0%, respectively, for the benchmark's fixed, hand-written injection. The red-teaming principles discovered against one victim also improve attacks against other victims, including a different model family, without additional feedback.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.30207 — 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 2026-10-17T07:28:26.836Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).