The rapid deployment of AI-driven autonomous agents in critical applications has increased the importance of secure and reliable inter-agent communication. This study presents a novel framework that integrates Multi-Agent Reinforcement Learning (MARL) with secure communication protocols to enhance trust, privacy, and resilience in AI-agentic interactions. The proposed approach leverages decentralized training and policy sharing, while employing cryptographic techniques—such as end-to-end encryption and key exchange mechanisms—to safeguard information flow among agents. Experimental results in simulated environments demonstrate that our method not only maintains competitive task performance but also significantly reduces vulnerabilities to eavesdropping, message tampering, and adversarial manipulation. By aligning MARL coordination strategies with robust security mechanisms, this research provides a scalable and adaptive foundation for deploying multi-agent systems in high-stakes domains, including autonomous robotics, distributed IoT networks, and secure collaborative decision-making platforms.
Musunuru et al. (Thu,) studied this question.