AI-Enabled Intrusion Detection in Wireless Sensor Networks for Critical Infrastructure

Abstract

Wireless Sensor Networks (WSNs) are essential in the surveillance and control of essential infrastructure applications such as smart grids, medical systems, and industrial automation. Nevertheless, they are extremely susceptible to various cyber threats due to their decentralized design and little computational and energy capabilities. In order to overcome these issues, this paper suggests a smart intrusion detection framework by Hybrid Adaptive Graph Attention Reinforcement Learning (HAGARL) approach. The proposed model combines graph-based learning to identify the spatial relationship between sensor nodes, a temporal reconstruction model to detect irregularities in the sequence traffic data, and an adaptive learning model to dynamically improve detection strategies. This combination allows the efficient analysis of structural and temporal network behaviors at the same time without consuming excessive resources in resource-constrained environments. It has been established experimentally that the proposed framework attains a detection rate of 99.3% and a false positive rate of 1.8% leading to better performance than some of the current methods. These results demonstrate that the hybrid model is effective towards improving the security, reliability, and operational efficiency of WSN-based critical infrastructure systems.