AN INTELLIGENT DATA-DRIVEN MODEL TO SECURE INTRAVEHICLE COMMUNICATIONS BASED ON MACHINE LEARNING

Abstract

Modern vehicles rely on Electronic Control Units (ECUs) and the Controller Area Network (CAN) bus for communication among vehicle components. However, the CAN protocol lacks inherent security mechanisms such as authentication, encryption, and access control, making intravehicle communication vulnerable to message injection, spoofing, replay, and Denial-of-Service (DoS) attacks. This project proposes an intelligent data-driven intrusion detection model based on Machine Learning to identify abnormal CAN bus communication and enhance vehicle cybersecurity. The proposed framework uses Support Vector Machine (SVM) for classifying CAN messages as normal or malicious, while Social Spider Optimization (SSO) is employed to optimize SVM parameters and improve detection performance. CAN bus data are preprocessed and communication features are extracted before model training and evaluation. The system compares KNN, conventional SVM, and SSO-SVM models using performance measures such as accuracy, hit rate, miss rate, false rate, and classification rate, supporting effective anomaly detection for secure intravehicle communication efficiently. Keywords: Intravehicle Communication, Support Vector Machin, Social Spider Optimization, Anomaly Detection, Vehicle Cybersecurity