Classification of intrusion using Multi-Stream Generative Adversarial Networks in Wireless Sensor Networks

Wireless Sensor Networks (WSNs) are extensively utilized in environmental monitoring, military applications and queue tracking. The short lifespan of sensor nodes are affected by battery depletion, particularly close to the sink node, frequently limits their efficacy and causes energy-holes and node loss. In this paper, an Optimal Cluster Routing Formation and Classification of Intrusion utilizing Multi-Stream Generative Adversarial Networks (OCRF-WSN-MSGAN) is proposed. The proposed approach aims to decrease energy consumption and extend node lifespan while improving intrusion classification. It selects the Cluster Heads (CH) using a fitness function derived from Multi-Stream Generative Adversarial Networks (MSGAN). The system initializes with a model for energy and path selection using Adaptive Tunicate Swarm Algorithm (ATSOA). The MSGAN classifies data as normal or abnormal, with its weight parameters optimized by Mayfly Optimization Algorithm (MOA). The proposed OCRF-WSN-MSGAN model is implemented in MATLAB and compared with existing methods under the performance metrics, like network lifetime, number of alive nodes, throughput, packet delivery ratio (PDR) and delay. The proposed OCRF-WSN-MSGAN method attains 20.76%, 26.45%, and 31.79% greater network lifetime, and 19.37%, 16.10%, and 31.23% lesser energy consumption compared with the existing techniques like the construction of an ideal cluster, trusted path for routing and classifying intrusions in WSN using machine learning (OCRF-WSN-RBPDT), combined glow-worm swarm and ant colony optimization for energy effective clustering in WSN(OCRF-WSN-ACI-GSO) and an energy-efficient routing technique utilizing the exponentially-ant lion whale optimization method in WSN(OCRF-WSN-EALWO) respectively.