An IoT-Based Deep-Learning Architecture to Secure Automated Electric Vehicles Against Cyberattacks and Data Loss
Shimaa Bergies, Tawfiq M. Aljohani, Shun‐Feng Su, Mahmoud Elsisi
Abstract
In the realm of modern transportation, automated electric vehicles (AEVs) assume a seminal role in realizing the vision of intelligent and electrified mobility. The advancement of AEVs hinges on the utilization of smart Internet of Things (IoT) devices as indispensable components to propel their evolution. These devices not only amplify the operational capabilities of AEVs but also underpin their security in the face of escalating cyber threats. In this regard, this study proposes a novel IoT architectural paradigm, encompassing integration of model predictive control (MPC) and deep neural network (DNN) frameworks. The proposed architecture aims to enhance AEV performance, empowering them to counteract the disruptive impact of erroneous data intrusions that result from cyber breaches. To ensure the timely identification of potential threats without compromising privacy considerations, this study augments the framework to encompass trajectory prediction. This extension is achieved through dynamic programming (DP) to craft effective control strategies governing AEV motions, conjoined with DNNs adept in discerning deviations from projected behavioral norms within AEVs’ control signals. This cohesive symbiosis propels the expeditious detection of anomalies indicative of potential security breaches. As data privacy remains a paramount consideration, this work employs Homomorphic Encryption, enabling anomaly score computation on encrypted data, thereby upholding privacy standards. Various test scenarios are conducted to emphasize the effectiveness of the proposed IoT architecture with MPC, DP, and DNN to improve the performance of the AEV. The results attest that the proposed approach can tackle cyberattacks and data loss effectively which enhances the production process and decision making.