Modern vehicles rely heavily on Controller Area Network (CAN) communication for interaction among Electronic Control Units (ECUs). However, CAN lack. s inherent security mechanisms, making it vulnerable to cyberattacks such as spoofing, replay, and denial-of-service. This paper proposes an Edge AI- driven framework integrated with centralized Federated Learning (FL) for predictive maintenance and intrusion detection in automotive systems. The proposed system employs two dedicated intelligent models: a predictive maintenance model that analyzes sensor data such as brake temperature, pressure, and vehicle speed to forecast potential failures, and a separate intrusion detection model based on a 1D Convolutional Neural Network (1D-CNN) to identify anomalous CAN message patterns. Both models are deployed in a decentralized manner at the edge level within individual vehicles, enabling real-time processing with minimal latency. In contrast, the Federated Learning process follows a centralized architecture, where locally trained model updates from multiple vehicles are transmitted to a central aggregation server. The server performs weighted parameters averaging to generate a global model, which is then redistributed to all vehicles to improve overall accuracy while preserving data privacy. Experimental results indicate that the proposed system achieves approximately 98% detection accuracy with latency below 10 ms, demonstrating its effectiveness for real-time automotive applications

This paper presented a Federated Learning-based intrusion detection and predictive maintenance system for automotive ECUs. The proposed framework achieves high accuracy, low latency, and strong privacy preservation, making it suitable for next-generation intelligent vehicles.

[1]Y. Zhang et al., “A two-stage federated learning- based transformer intrusion detection system for CAN,” Cybersecurity, 2025.Available: https://link.springer.com/article/10. 1186/s42400-024- 00329-2 [5] S. Ghosh, A. S. M. M. Jameel, and A. El Gamal, “FetFIDS: A Feature Embedding Attention Based Federated Network Intrusion Detection Algorithm,” arXiv preprint, Aug. 2025. Available: https://arxiv.org/abs/2508.09056 [6]A. A. Mazroa, “FORT-IDS: A Federated, optimized, robust and trustworthy intrusion detection system for IIoT security,”Sci.Rep.,2026. Available: https://doi.org/10.1038/s41598- 025-31025-x [7]P. Narang et al., “FedLiTeCAN: A Federated Lightweight Transformer for Real-Time CAN Intrusion Detection,” arXiv preprint, Dec. 2025. Available: https://arxiv.org/pdf/2512.24088 [8]M. Devi et al., “Federated Learning-Enabled Lightweight Intrusion DetectionSystems,” Comput. Secur., 2025. Available: https://www.sciencedirect.com/scie nce/article/pii/S266730532500079 1 [9]N. Soomro et al., “SecureDyn- FL: A Robust Privacy- Preserving Federated Framework for IDS,” arXiv preprint,Jan.2026. Available: https://arxiv.org/abs/2601.06466

© 2025 International Journal of Engineering and Techniques (IJET).

Digital Object Identifier (DOI)

IJET assigns a unique DOI to every accepted article via Zenodo for persistent linking and citation. Your article’s DOI: https://doi.org/{{doi}}

Open DOI About Zenodo DOI

Close