TRUST-AWARE BLOCKCHAIN-ASSISTED MULTI-AGENT DRL FOR SECURE AND SCALABLE ROUTING IN IOT-ENABLED MANETS

Abstract

This paper proposes a trust-aware blockchain-assisted multi- agent deep reinforcement learning (MADRL) framework for secure and scalable routing in IoT-enabled Mobile Ad Hoc Networks (MANETs) under adversarial environments. The proposed framework integrates hierarchical blockchain- based trust management with distributed Double Deep Q- Network (DDQN)-based MADRL to enable adaptive and security-aware routing decisions. A trust-aware reward shaping mechanism is introduced, where blockchain-derived trust values dynamically influence policy learning through smart-contract-assisted reward adjustment. In addition, a lightweight hierarchical blockchain architecture is designed to reduce synchronization overhead while preserving distributed trust consistency in dynamic MANET environments. Extensive simulations were conducted using MANET scenarios with 50-500 mobile nodes under blackhole, Sybil, and jamming attacks. Experimental results show that the proposed framework improves packet delivery ratio (PDR) by 10-15%, increases malicious node detection accuracy by up to 20%, and reduces end-to-end delay by 10- 18% compared with conventional routing schemes. The results confirm that the proposed framework achieves an effective balance among security, reliability, and scalability for next-generation IoT-enabled MANET systems. Keywords: IoT-enabled MANETs; Secure routing; Blockchain; Multi-agent deep reinforcement learning (MADRL); Double Deep Q-Network (DDQN).