Deep-Learning-Powered Cyber-Attacks Mitigation Strategy in the EV Charging Infrastructure
Manoj Basnet, Mohd. Hasan Ali
Abstract
The Electric Vehicle Charging Station (EVCS) migrates the vulnerabilities of the incumbents’ technologies, including communication, control, supply chain, software, and humans. The threat actors can exploit the vulnerabilities to freeze, disrupt, damage, and congest the charging services. State-of-the-art technologies are evolving for cyber threat detection and isolation in EVCS at the network and physical levels. However, the current literature lacks the proper mitigation technologies to deal with cyber-enabled physical attacks and physically-enabled cyberattacks at EVCS. To overcome these limitations, we propose the Long-Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) based mitigation approaches to deal with the Advanced Persistent Threats (APT) attacks at standalone EVCS. The performance measures indicate effective mitigation results by the proposed models. The results also indicate the superiority of LSTM-based automated mitigation in terms of adaptation to system dynamics and control actions compared to GRU-based mitigation.