IoT based Real Time Anomaly Detection in Automotive ECUs Using Machine Learning

Electronic control units (ECU s) play a vital role in modern vehicles, but their increased use opens vehicles to cyber-attacks that threaten safety and privacy. This research proposes a novel approach for real-time detection and mitigation of ECU attacks, leveraging machine learning (ML) and IoT communication protocols. The methodology involves these key stages: data collection, data preprocessing, anomaly detection, and IoT -enabled alerting. Simulation of a Controller Area Network (CAN) facilitates the capture of ECU communication packets. Data preprocessing involves normalization, calculation of message reception intervals, and extraction of salient features. This work focuses on slope analysis for anomaly detection. A significant deviation in slope from a baseline pattern could indicate a potential attack. Upon detection, an alert is generated and transmitted through IoT protocols (e.g., MQTT), facilitating real-time notification. A proof-of-concept demonstrates the solution's effectiveness in detecting real-world attack scenarios, including introducing a malicious ECU to disrupt communication. Combining ML and IoT integration, this approach achieves accurate detection. This work advances automotive cyber-security by providing a real-time, reliable ECU attack detection and alerting solution, promoting safer driving experiences.