Intrusion Detection System in CAN-BUS Vehicle Networks Using Machine Learning

Abstract

As vehicles become more connected and software-rich, securing communications inside vehicles is increasingly important. The Controller Area Network (CAN-BUS) is simple and efficient, which is why it remains the dominant vehicle network. That simplicity comes at a cost: the protocol lacks built-in authentication or encryption, leaving it exposed to injection, flooding and impersonation attacks. Here we describe a practical, real-time Intrusion Detection System (IDS) for CAN-BUS that combines supervised machine learning with features useful for deployment. The core detector is a Random Forest trained on the OCS-Lab dataset; it distinguishes benign traffic from DoS, fuzzy and impersonation attacks and achieves a validation accuracy of 85.37%. Importantly, the IDS is more than a model: it includes a retraining dashboard, a timeline for investigation, live packet classification and automated email alerts so operators can react quickly and with context.