An adaptive, energy-efficient and secure routing protocol for zone-related mobile Ad-hoc networks using reinforcement learning.

The rapid surge of Mobile Ad Hoc Networks (MANETs) stimulates the need for adaptive, intelligent, and secure routing mechanisms to ensure seamless communication in dynamic environments. Traditional routing protocols are battling with security threats such as wormhole attacks that disrupt routing and degrade network performance. To address these challenges, this study outlines a state-of-the-art technique, the Reinforcement Learning-Based Secure Routing Protocol (RLSRP), which leverages adaptive k-hop clustering and deep Q-Networks (DQN) to fine-tune routing decisions dynamically while mitigating security risks. RLSRP unwaveringly measures network condition by evaluating latency variations and anomalies to spot suspicious nodes, thereby enhancing route stability. The protocol implements zone-related clustering where nodes within each zone collaborate to optimise routing paths based on real-time conditions, ensuring energy-efficient communication. Current research investigated deep reinforcement learning methodologies to improve security in zone-related MANETs and ensure efficient data routing in large-scale environments. A detailed simulation-based evaluation depicts the potency of the proposed RLSRP model when compared with other reinforcement learning-based routing protocols. Using a large-scale setup-scaling up to 10 million nodes-with Dask and TensorFlow, the results show that RLSRP consistently outperforms FSSAM, Cluster-RL, and Reputation-based Q-learning in terms of Packet Delivery Ratio (exceeding 99%), reduced latency, and improved energy efficiency. These findings reported RLSRP as a secure and scalable solution for practical MANET routing.