Impact of Class Balancing on Intrusion Detection System for WSN-BFSF Dataset

Wireless Sensor Networks (WSNs) are crucial to various applications within the Internet of Things (IoT), Industry 4.0, Cyber-Physical Systems (CPS), and smart manufacturing. The rising incidences of cyber-attacks highlight the critical need for advanced security measures in WSNs. This study aims to develop a robust Intrusion Detection System (IDS) that leverages Machine Learning (ML) techniques to safeguard WSNs from network intrusions. In 2023, the WSN-BFSF dataset, which includes data on normal traffic and blackhole, selective forwarding, and flooding attacks, became available. As a crucial defense for securing WSNs, IDS detects and mitigates attacks by constantly monitoring network traffic for unusual activity, identifying specific attack patterns, and quickly isolating malicious nodes while updating routing protocols. This research proposes a novel IDS mechanism that employs the Synthetic Minority Oversampling Technique (SMOTE) for class balancing on the WSN-BFSF dataset. Class balancing enhances intrusion detection by reducing bias toward majority classes. Our new IDS model is compared to the original research that introduced the WSN-BFSF dataset. The results show significant performance improvements across various ML algorithms: Precision increased by 2.23–12.40%, Recall by 0.64–5.01%, F1-score by 3.63–9.12%, and overall accuracy by an average of 3.97%. These findings indicate that class balancing offers promising results for improving intrusion detection in WSNs.