Enhancing intrusion detection in wireless sensor networks using a Tabu search based optimized random forest

Intrusion detection in Wireless Sensor Networks (WSNs) is an emerging area of research, given their extensive use in sensitive fields like military surveillance, healthcare, environmental monitoring, and smart cities. However, WSNs face several security challenges due to their limited computational capabilities and energy constraints. Their deployment in open, unattended environments makes them especially vulnerable to threats like eavesdropping, interference, and jamming. To address this problem, Random Forest (RF) is a popular machine learning model. The RF model can be tweaked because of its multiple hyperparameters. Tuning these parameters manually is tedious, as the combinations will be exponential. This work presents an enhanced intrusion detection approach by integrating Tabu Search (TS) optimization with a RF classifier. As a result, TS will help RF automatically search optimal hyperparameters and improve the generalization ability. This work integrates the pros of TS with RF. The model was tested on three different datasets, i.e., (a) the WSN-DS dataset, (b) CICIDS 2017, and (c) the CIC-IoT 2023 dataset, which shows better results on different metrics like precision, recall, F1-score, Cohen's Kappa, and ROC AUC. Detection of Blackhole and Gray Hole attacks also improved, demonstrating the effectiveness of combining metaheuristic optimization with ensemble learning for stronger WSN security.