A Quantum-Resilient Federated DRL Framework for Secure Voice Communication in 6G-Enabled MANETs

The problem of secure voice communications in Mobile Ad-hoc Networks (MANETs) is gaining traction as 6G low-latency applications are introduced and the threat of a quantum adversary is growing. The classical encryption schemes are vulnerable to lattice attacks by Shor, and centralized learning schemes are unable to maintain privacy in decentralised military or disaster response MANETs. In this paper, we present Federated Deep Reinforcement Learning-Driven Post-Quantum Voice Encryption Framework (FedRL-PQVE) that incorporates lattice-based key generation, federated policy learning, and real-time adaptive encryption optimization to MANET routing state. DRA agent maximizes the strength of encryption, bit-allocation, and computational cost per hop without infringing the privacy of the user by means of decentralized model aggregation. It has been demonstrated to reduce latency by 28-42 percent, reduce the probability of quantum attack success by 50-70 percent and reduce energy consumption by 30 percent over classical systems. The framework provides a powerful and smart quantum-resistant voice communication application to 6G-enabled MANETs.