Metraipolis › archive › dc_4ffb3df01f0b

Securing Computer-Use Agents Against Branch Steering Attacks — arXiv:2610.03089

Source
arXiv
Original
https://arxiv.org/abs/2610.03089
License (as archived)
CC0-1.0
Archived
2026-10-05T22:50:26.320Z
SHA-256 of the archived text
45fb4b502d6b383536306e56b2dd128c1584f66747abb6e67e798f2e2d9c0c33
Stable URL of this page
https://metraipolis.com/documento/dc_4ffb3df01f0b

How to cite

Securing Computer-Use Agents Against Branch Steering Attacks — arXiv:2610.03089 — arXiv, https://arxiv.org/abs/2610.03089 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4ffb3df01f0b (Metraipolis archive dc_4ffb3df01f0b, sha256 45fb4b502d6b383536306e56b2dd128c1584f66747abb6e67e798f2e2d9c0c33)
BibTeX
@misc{dc4ffb3df01f0b,
  title = {Securing Computer-Use Agents Against Branch Steering Attacks — arXiv:2610.03089},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.03089},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4ffb3df01f0b; sha256 45fb4b502d6b383536306e56b2dd128c1584f66747abb6e67e798f2e2d9c0c33},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.03089. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Modern Computer Use Agents (CUAs) directly interact with graphical user interfaces and execute third-party web tools, exposing them to indirect prompt injection across every rendered page and tool response. While the Dual-LLM pattern is the primary system-level architecture offering formal security guarantees - using an isolated Planner LLM (P-LLM) to fix execution paths before processing untrusted inputs via a Quarantined LLM (Q-LLM) - these guarantees break down in graphical environments. Because CUA interaction is inherently dynamic, plans cannot remain data-independent; they must branch based on anticipated runtime web content - covering all possible cases the agent may encounter. This exposes agents to branch steering attacks, where an adversary crafts untrusted data to coerce a CUA down a hazardous, pre-approved branch without injecting explicit instructions. We systematically study branch steering attacks and introduce STEER-Bench (101 tasks across 9 domains), showing high attack success against both standard (94.4%) and vanilla Dual-LLM (89.5%) CUAs. We then propose COBRA, an architecture that pairs trusted branching plans with ahead-of-time capability constraints, strictly bounding the parameters and destinations each branch may execute. On STEER-Bench, COBRA reduces attack success to 0% while retaining 97% benign utility.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.03089 — 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 2026-10-05T22:45:55.826Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).