Two-Pronged Intrusion Detection System for MANET

Mobile Ad hoc Networks (MANETs) provide a flexible wireless infrastructure for various applications. Still, their decentralized and dynamic nature makes them susceptible to security threats. The topology of a MANET changes frequently due to node mobility. This dynamic nature can result in frequent network disruptions, packet losses, and increased latency. This paper introduces a Two-Pronged Intrusion Detection System (TP-IDS) designed for MANET, focusing on blackhole, hello flooding, and UDP flooding attacks. These attacks take advantage of the dynamic and self-organizing features of MANETs, making the network perform poorly. A dataset with eight features is created, capturing mobility, received, sent, and dropped packets, forwarding difference, energy, and transmission power. We have used SMOTE to overcome the class imbalance. Hyperparameter Optimization is performed using Bayesian Optimization and Optuna across Decision Tree, Random Forest, and XGBoost classifiers showcasing strengths and weaknesses for optimal selection. We conclude that the XGBoost classifier with Grid-initialized Bayesian hyperparameter optimization provides the best overall performance to create a robust TP-IDS.