SentinelNet: Safeguarding Multi-Agent Collaboration Through Credit-Based Dynamic Threat Detection

Malicious agents pose significant threats to the reliability and decision-making capabilities of Multi-Agent Systems (MAS) powered by Large Language Models (LLMs). Existing defenses often fall short due to reactive designs or centralized architectures which may introduce single points of failure. To address these challenges, we propose SentinelNet, the first decentralized framework for proactively detecting and mitigating malicious behaviors in multi-agent collaboration. SentinelNet equips each agent with a credit-based detector trained via contrastive learning on augmented adversarial debate trajectories, enabling autonomous evaluation of message credibility and dynamic neighbor ranking via bottom-k elimination to suppress malicious communications. To overcome the scarcity of attack data, it generates adversarial trajectories simulating diverse threats, ensuring robust training. Experiments on MAS benchmarks show SentinelNet achieves near-perfect detection of malicious agents, close to 100% within two debate rounds, and recovers 95% of system accuracy from compromised baselines. By exhibiting strong generalizability across domains and attack patterns, SentinelNet establishes a novel paradigm for safeguarding collaborative MAS.

Key Contributions

• SentinelNet: the first decentralized framework for proactively detecting malicious agents in LLM multi-agent systems using contrastive learning-based credit scoring per agent
• Bottom-k dynamic neighbor elimination that suppresses malicious communications without centralized coordination, avoiding single points of failure
• Adversarial trajectory generation method to synthesize diverse attack data and overcome training data scarcity for malicious-agent detectors

🛡️ Threat Analysis

Details

Similar Papers