Metraipolis › archive › dc_d4455db06c05
Vulnerability Analysis for Safe Multi-Agent Reinforcement — arXiv:2610.27293
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.27293
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-21T07:06:24.576Z
- SHA-256 of the archived text
f28e33d7a10919ceaa9659f5d83398dd2b17be48e24abee30f3898435160d034- Stable URL of this page
https://metraipolis.com/documento/dc_d4455db06c05
How to cite
Vulnerability Analysis for Safe Multi-Agent Reinforcement — arXiv:2610.27293 — arXiv (en), https://arxiv.org/abs/2610.27293 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d4455db06c05 (Metraipolis archive dc_d4455db06c05, sha256 f28e33d7a10919ceaa9659f5d83398dd2b17be48e24abee30f3898435160d034)
BibTeX
@misc{dcd4455db06c05,
title = {Vulnerability Analysis for Safe Multi-Agent Reinforcement — arXiv:2610.27293},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.27293},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d4455db06c05; sha256 f28e33d7a10919ceaa9659f5d83398dd2b17be48e24abee30f3898435160d034},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.27293. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Safe Multi-Agent Reinforcement Learning (SMARL) aims to learn cooperative control policies under prescribed safety constraints. Al- though existing SMARL methods focus on enforcing safety during policy optimization, their vulnerability remains largely unexplored. This paper investigates inference-time observation attacks against SMARL, where an adversary perturbs the observations of a subset of agents, thereby altering the actions produced by their policies. We propose MinMax Attack, an attack framework that jointly selects adversarial target actions for compromised agents while accounting for the Cooperative Resistance of benign agents. We further derive a bound on the induced safety cost and characterize its relation- ship with the Compromised-Agent Budget and the Perturbation Budget. Experiments across multiple safety-critical multi-agent benchmarks and representative state-of-the-art SMARL algorithms show that MinMax Attack induces substantially higher safety costs than existing attack baselines while preserving task reward, thereby reducing its detectability and achieving both attack effectiveness and stealthiness. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.27293 — 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 2026-10-21T07:05:43.392Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).