Personalized Federated Learning for Automotive Intrusion Detection Systems
Kabid Hassan Shibly, Md Delwar Hossain, Hiroyuki Inoue, Yuzo Taenaka, Youki Kadobayashi
Abstract
In connected cars, the Controller Area Network (CAN) bus communication is the central connectivity and communication system for electronic control units (ECUs). Although the CAN bus is the central communication system for most cars, it lacks basic security features, i.e., authentication and encryption. Consequently, an attacker may compromise the CAN bus system effortlessly with even free attacking tools. In case of an attacker succeeds in compromising the ECUs, they can take control and stop the engine, disable the brakes, turn the lights on/off, etc., which makes the questions concerning the transformation of modern cars and safe driving. In this study, we propose a Personalized Federated learning-based Intrusion Detection System that ensures effective, secure training procedures without sharing any sort of data. In our research, we contemplate Supervised and Unsupervised Federated Learning to observe the behavior of CAN bus intrusion data. Our experiment result demonstrates that the Federated Learning-based supervised classifier effectively detects the CAN bus attacks, with accuracy of 99.98%.