AI-Powered Secure and Privacy-Aware Maintenance Prediction Scheme for Autonomous Vehicles Using Hardware Acceleration

With advancements in Internet of Things (IoT) technologies for Intelligent Transportation Systems (ITS), gathered vehicle data can provide insights into emerging vehicular phenomena and help the continued enhancement of creative and efficient vehicular systems. Overall, improvements to ITS have had a significant influence on society. Predictive maintenance will discover faults within the vehicle and offer early warnings to avert failure by using data collected from car sensors and maintenance models built from prior vehicle repairs. The primary goal of this study is to develop a secure, privacy-preserving, and continuous data-gathering strategy for predictive maintenance, utilizing a K-Nearest Neighbor (KNN) aggregation over an encrypted data scheme and a Neural Network (NN) prediction model. In this suggested approach, data will be exchanged among vehicles, fog nodes, and a cloud server. The vehicle’s sensors will produce a sensory data report and transmit it to the fog nodes after encryption to preserve the vehicle’s user privacy. The encrypted information will be aggregated through the fog nodes before being transferred to the cloud server. Finally, the cloud server will issue maintenance details to fog nodes and vehicles using an NN predictive model. Furthermore, the proposed scheme’s capability for implementation on hardware is comprehensively examined and evaluated. Our security and privacy evaluation shows that the scheme can achieve our design goals. Additionally, our performance evaluation shows that our scheme has low computation and communication overheads compared with the existing techniques. Especially for the NN-based model that achieves 100% accuracy, 100% precision, 97.5% recall, and 98.7% F1-score through the software tests; at the same time, it performs with good utilization values on the FPGA board.