Multi-Agent System Security

Follow this topic Ask about this topic

Multi-agent system security examines how agents can attack, mislead, or compromise one another, and how local failures can spread through communication, shared memory, delegated tools, or coordinated decisions. Work studies threats such as prompt injection, collusion, Byzantine behavior, memory poisoning, information leakage, and cascading errors, using settings from language-model workflows and debate to multi-agent reinforcement learning and robot swarms. Defenses include trust and reputation systems, provenance and authorization controls, runtime monitoring, semantic information-flow checks, and cryptographic protocols; benchmarks increasingly test different architectures, interaction paths, and long-horizon tasks. A central challenge is that safeguards effective for individual agents may not compose at system level, while stronger controls can reduce task performance or miss attacks that emerge only through multi-agent interaction.

Latest papers

181 papers29 papers in the last four weeks, up 12% on the four weeks before. 0.2% of all new papers.

SUB-PLAY: Adversarial Policies against Partially Observed Multi-Agent Reinforcement Learning Systems
Feb 6, 2024Oubo Ma, Yuwen Pu, Linkang Du +5Adversarial Attacks Partially Observable RL
Zhejiang University · Xi’an Jiaotong University · Laboratory for Big Data and Decision +2

Feb 6, 2024J/K move · Enter open · S save

SUB-PLAY: Adversarial Policies against Partially Observed Multi-Agent Reinforcement Learning Systems

Oubo Ma, Yuwen Pu, Linkang Du, Yang Dai+4

Zhejiang University · Xi’an Jiaotong University · Laboratory for Big Data and Decision+2

Adversarial Attacks Partially Observable RL Black-Box Adversarial Attacks Multi-Agent System Security Multi-Agent Reinforcement Learning

Open paper AskPDF arXiv

Recent advancements in multi-agent reinforcement learning (MARL) have opened up vast application prospects, such as swarm control of drones, collaborative manipulation by robotic arms, and multi-target encirclement. However, potential security threats during the MARL deployment need more attention and thorough investigation. Recent research reveals that attackers can rapidly exploit the victim's vulnerabilities, generating adversarial policies that result in the failure of specific tasks. For instance, reducing the winning rate of a superhuman-level Go AI to around 20%. Existing studies predominantly focus on two-player competitive environments, assuming attackers possess complete global state observation. In this study, we unveil, for the first time, the capability of attackers to generate adversarial policies even when restricted to partial observations of the victims in multi-agent competitive environments. Specifically, we propose a novel black-box attack (SUB-PLAY) that incorporates the concept of constructing multiple subgames to mitigate the impact of partial observability and suggests sharing transitions among subpolicies to improve attackers' exploitative ability. Extensive evaluations demonstrate the effectiveness of SUB-PLAY under three typical partial observability limitations. Visualization results indicate that adversarial policies induce significantly different activations of the victims' policy networks. Furthermore, we evaluate three potential defenses aimed at exploring ways to mitigate security threats posed by adversarial policies, providing constructive recommendations for deploying MARL in competitive environments.