Intrusion detection using graph neural network and Lyapunov optimization in wireless sensor network

Sensor nodes deployed in a remote location are vulnerable to various attack. An intruder can easily capture and tamper with sensor nodes deployed in a remote location. As a result, intrusion detection is crucial task in the field of wireless sensor network. In this work, we propose an intrusion detection approach for WSN. In our method,we are using Graph Neural Network and Lyapunov optimization. In the training phase, we train graph data using GNN. We are using Lyapunov optimization to adjust weights of the synapses connecting two neurons to an optimum value. Here we used AWID datasets to train and test GNN. Lyapunov optimization is used to compute loss in GNN and adjust weight accordingly to minimize loss. We show test results of our method using performance matrices, namely, Accuracy, Sensitivity, Precision, F1 Score. Comparison with existing work showed that our method gives better detection accuracy.