Federated Intrusion Detection for Smart Vehicles

Cybersecurity threats increasingly endanger networked systems consisting of multiple nodes, such as Internet of Things (IoT) systems, Supply Chains and Smart Cars, due to the emergence of new types of zero-day attacks and the vulnerability of IoT devices. While earlier research shows that Machine Learning (ML)-based Intrusion Detection Systems (IDSs) enhance security, the effectiveness and practicality of ML-based IDSs with large training datasets, are often challenged in distributed systems, where each node must ensure the confidentiality of its local data, despite similar attacks against other nodes in the system.Thus, this paper presents iDAF, a novel Decentralized and Asynchronous IDS using Online Self-Supervised Federated Learning, which improves overall security by providing collaborative data confidentiality-preserving learning between all nodes. iDAF enables self-supervised learning without human intervention, eliminates the need for large labelled data, and avoids longlasting local training sessions.We evaluate iDAF’s ability to detect DoS and DDoS attacks in Collaborating Connected Smart Cars using open-access datasets and compare its performance against five other methods. The results show that iDAF delivers high intrusion detection accuracy with an acceptable computation time and low communication overhead. Thus, iDAF system can be an enabler for collaborative security tools in connected vehicles.