ML-PLSR: AN INTEGRATED LIGHTWEIGHT PHYSICAL LAYER SECURED ROUTING ALGORITHM FOR MULTI-HOP WIRELESS NETWORKS INSPIRED BY RANDOM FOREST AND CNN-LSTM

Abstract

Multi-hop wireless networks, critical to Intelligent Transport Systems, IoT, and industrial automation, face significant security threats like DDoS, jamming, and eavesdropping due to their open and dynamic nature. Existing frameworks often lack real-time adaptability and scalability. This paper introduces ML-PLSR (Machine Learning Based Physical Layer Secured Routing Algorithm), a lightweight, integrated framework for real-time attack detection, intelligent routing, and secure data transmission. Utilizing Random Forest-inspired decision trees for detecting DDoS and jamming via live metrics (PDR, SINR), a Convolutional Neural Networks and Long Short-Term Memory (CNN-LSTM) inspired model for reliable forwarder selection based on spatio-temporal features, and ECC for end-to-end encryption, ML-PLSR ensures robust security and performance. Implemented in NS2 and evaluated across WSN, VANET, and Zigbee, it achieves up to 26.65% higher secrecy capacity, over 92.8% packet delivery ratio, and 98% detection accuracy, offering an efficient solution for securing next-generation wireless networks.