Balancing CICIoV2024 Dataset with RUS for Improved IoV Attack Detection

This study addresses the cybersecurity challenges within the Internet of Vehicles (IoV) by exploring the efficacy of Random Under-Sampling (RUS) in balancing the class distribution of the CICIoV2024 dataset for improved intrusion detection. IoV technology connects vehicles to digital infrastructure, fostering communication and enhancing safety but is simultaneously vulnerable to cyber threats such as Denial of Service (DoS) and spoofing attacks. This research employed RUS to mitigate data imbalance within the CICIoV2024 dataset, which often impedes effective threat detection in machine learning models. Four machine learning classifiers Random Forest, AdaBoost, Gradient Boosting, and XGBoost were evaluated on both imbalanced and balanced datasets to compare their performance. Results demonstrated that RUS significantly enhances model accuracy, precision, recall, and F1-score, reaching perfect scores across all classifiers post-balancing. Additionally, RUS contributed to substantial reductions in training and testing times, thereby boosting computational efficiency. These findings underscore the potential of RUS in addressing data imbalance in IoV cybersecurity, establishing a foundation for future research aimed at safeguarding IoV systems against evolving cyber threats.