Enhanced Intrusion Detection in Wireless Sensor Networks: A Voting and Particle Swarm Optimization Approach

Wireless sensor networks (WSNs) are of great importance in advanced technologies in the modern developing world as they are widely used in many applications. Although encryption and authentication protocols in WSNs are widely adopted, there are continuous expectations of new sophisticated attack strategies that are spreading, which pose high risks and threats. This work addresses these challenges using Intrusion Detection Systems (IDS) that will be supported by machine learning algorithms. Specifically, it explores the WSN-DS dataset that includes network activities and attack types. In addition, it initializes the importance of feature selection and proposes data balancing methods such as Synthetic Minority Over-sampling Technique SMOTE, it is a statistical method for correcting class imbalance in datasets by creating synthetic samples for the minority class to enhance the detection rate. To increase the performance of the machine learning model, we use a voting classifier consisting of Random Forest (RF), K-Nearest Neighbors (KNN) and Extreme Gradient Boosting (XGBoost) and optimize them using Particle Swarm Optimization (PSO) by selecting the best hyperparameters. This ensemble method aims to maximize the performance of the used ensemble models. The proposed model is demonstrated by improving the PSO voting classifier model which gives a remarkable accuracy of 99.99% making it a highly effective approach in predicting and filtering different types of attacks in wireless sensor networks. This work will help in promoting further research in protecting wireless sensor networks from invasive intrusions in various applications to ensure security in communication networks.