← 研踪 PaperTrailarXiv:2508.09275 · v3
Constrained Black-Box Attacks Against Cooperative Multi-Agent Reinforcement Learning
首次提交 2025-08-13 02:31
版本更新 2026-11-04 05:19
雷达观测 2026-11-05 11:11
摘要
Collaborative multi-agent reinforcement learning (MARL) is increasingly deployed in critical domains, yet its vulnerability to adversarial perturbations has not been thoroughly investigated. Most existing adversarial attacks against MARL rely on optimistic assumptions, such as access to policy weights or the ability to train surrogate policies. We investigate whether adversaries with limited capabilities remain effective against MARL agents. In particular, we study two restricted attack settings that are overlooked in the literature: (i) adversaries with access only to agent observations and (ii) adversaries with no access whatsoever. Our approach generates perturbations that intentionally misalign how victim agents perceive their environment, preventing them from collaborating. We conduct extensive experiments across a diverse set of 22 environments. Our results demonstrate both the effectiveness and sample efficiency of our attacks.
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v3 · 2026-11-04 05:19 · 最新观测版本
Constrained Black-Box Attacks Against Cooperative Multi-Agent Reinforcement Learning
Collaborative multi-agent reinforcement learning (MARL) is increasingly deployed in critical domains, yet its vulnerability to adversarial perturbations has not been thoroughly investigated. Most existing adversarial attacks against MARL rely on optimistic assumptions, such as access to policy weights or the ability to train surrogate policies. We investigate whether adversaries with limited capabilities remain effective against MARL agents. In particular, we study two restricted attack settings that are overlooked in the literature: (i) adversaries with access only to agent observations and (ii) adversaries with no access whatsoever. Our approach generates perturbations that intentionally misalign how victim agents perceive their environment, preventing them from collaborating. We conduct extensive experiments across a diverse set of 22 environments. Our results demonstrate both the effectiveness and sample efficiency of our attacks.
查看此版本解读 →查看此版本原文 ↗