Heuristic-Optimized Machine Learning Models for Enhanced Attack Detection in Diverse Wireless Sensor Network Protocols

Abstract

Ensuring robust security in Wireless Sensor Networks (WSNs) is vital due to their susceptibility to various types of attacks that can compromise data integrity and network performance. This study investigates the effectiveness of machine learning models for intrusion detection across multiple WSN protocols, including TEEN, HEED, LEACH, and CFA-LEACH. Traditional classifiers such as K-Nearest Neighbors, Random Forest, Naïve Bayes, Decision Tree, and Multi-Layer Perceptron are evaluated and further enhanced through Particle Swarm Optimization (PSO) to improve detection accuracy. The research emphasizes comparative analysis across key performance metrics, demonstrating that PSO-based models offer significant improvements in classification performance. The study also explores protocol-specific attack prevalence, offering insights into the varying vulnerabilities of each protocol. The findings support the integration of swarm intelligence with machine learning as a promising approach for enhancing security and resilience in WSN environments.