Leveraging Artificial Neural Networks and LightGBM for Enhanced Intrusion Detection in Automotive Systems.

Connected components in automobiles face cybersecurity threats if not adequately protected. The Controller Area Network (CAN), the primary communication protocol in vehicular systems, lacks essential security mechanisms, necessitating heightened cybersecurity awareness in the automotive industry. This paper presents a methodology to enhance cybersecurity within the CAN network, including a comparative study with benchmark results. Our research focuses on improving multiclass Intrusion Detection Systems (IDSs) for the CAN network. We introduce a hybrid deep learning model tailored for precise multiclass classification of attacks, an Artificial Neural Network (ANN)-based Light Gradient Boosting Machine (LightGBM) IDS that incorporates a comprehensive set of features. The evaluation on diverse cyber-attacks from the Car-Hacking dataset demonstrates the model's effectiveness. This study contributes to the field by offering a brief assessment of current state-of-the-art methods and presenting a novel approach to bolster cybersecurity. The proposed IDS serves as a valuable reference for ensuring the cybersecurity of automotive systems. [ABSTRACT FROM AUTHOR]