Deep Learning -Powered Computational Intelligence for Cyber-Attacks Detection and Mitigation in 5G-Enabled Electric Vehicle Charging Station

An electric vehicle charging station (EVCS) infrastructure is the backbone of transportation electrification. However, the EVCS has various cyber-attack vulnerabilities in software, hardware, supply chain, and incumbent legacy technologies such as network, communication, and control. Therefore, proactively monitoring, detecting, and defending against these attacks is very important. The state-of-the-art approaches are not agile and intelligent enough to detect, mitigate, and defend against various cyber-physical attacks in the EVCS system. To overcome these limitations, this dissertation primarily designs, develops, implements, and tests the data-driven deep learning-powered computational intelligence to detect and mitigate cyber-physical attacks at the network and physical layers of 5G-enabled EVCS infrastructure. Also, the 5G slicing application to ensure the security and service level agreement (SLA) in the EVCS ecosystem has been studied. Various cyber-attacks such as distributed denial of services (DDoS), False data injection (FDI), advanced persistent threats (APT), and ransomware attacks on the network in a standalone 5G-enabled EVCS environment have been considered. Mathematical models for the mentioned cyber-attacks have been developed. The impact of cyber-attacks on the EVCS operation has been analyzed. Various deep learning-powered intrusion detection systems have been proposed to detect attacks using local electrical and network fingerprints. Furthermore, a novel detection framework has been designed and developed to deal with ransomware threats in high-speed, high-dimensional, multimodal data and assets from eccentric stakeholders of the connected automated vehicle (CAV) ecosystem. To mitigate the adverse effects of cyber-attacks on EVCS controllers, novel data-driven digital clones based on Twin Delayed Deep Deterministic Policy Gradient (TD3) Deep Reinforcement Learning (DRL) has been developed. Also, various Bruteforce, Controller clones-based methods have been devised and tested to aid the defense and mitigation of the impact of the attacks of the EVCS operation. The performance of the proposed mitigation method has been compared with that of a benchmark Deep Deterministic Policy Gradient (DDPG)-based digital clones approach. Simulation results obtained from the Python, Matlab/Simulink, and NetSim software demonstrate that the cyber-attacks are disruptive and detrimental to the operation of EVCS. The proposed detection and mitigation methods are effective and perform better than the conventional and benchmark techniques for the 5G-enabled EVCS