IoV Cyberattack Detection via TabTransformer on CAN Bus Communications

Abstract

The Internet of Vehicles (IoV) represents a transformative evolution in transportation by linking vehicles through the Internet to improve communication and operational efficiency. This interconnected network relies heavily on the Controller Area Network (CAN) which facilitates communication among Electronic Control Units (ECUs) within vehicles. However, this increased connectivity also introduces vulnerabilities making IoV systems susceptible to various cyberattacks. This study paper focuses on identifying cyberattacks on IoV systems using advanced Machine Learning (ML) techniques. The TabularTransformer (TabTransformer) model was utilized to analyze the CICIoV2024 dataset. To address class imbalance inherent in the dataset, two resampling techniques were implemented: Synthetic Minority Oversampling - Edited Nearest Neighbor (SMOTEENN) and Synthetic Minority Oversampling-Tomek links (SMOTETomek). The comparative study evaluated the performance of the TabTransformer under different class distributions, including a baseline scenario. The results indicate that the SMOTEENN technique improved model performance, achieving a precision of 89%, recall of 85%, and F1-score of 86%.