Machine Learning-Enabled Cyber Attack Prediction and Mitigation for EV Charging Stations

TL;DR

This paper presents a cybersecurity framework for electric vehicle charging stations using threat modeling, attack scenario analysis, and probabilistic algorithms to predict and mitigate cyber threats, enhancing the safety of EV infrastructure.

Contribution

It introduces a novel cybersecurity approach combining STRIDE, attack trees, HMM, and POMCP algorithms for EVCS threat detection and mitigation.

Findings

Identified vulnerabilities in EV charging stations.

Developed probabilistic models for attack prediction.

Suggested mitigation strategies for cyber threats.

Abstract

Safe and reliable electric vehicle charging stations (EVCSs) have become imperative in an intelligent transportation infrastructure. Over the years, there has been a rapid increase in the deployment of EVCSs to address the upsurging charging demands. However, advances in information and communication technologies (ICT) have rendered this cyber-physical system (CPS) vulnerable to suffering cyber threats, thereby destabilizing the charging ecosystem and even the entire electric grid infrastructure. This paper develops an advanced cybersecurity framework, where STRIDE threat modeling is used to identify potential vulnerabilities in an EVCS. Further, the weighted attack defense tree approach is employed to create multiple attack scenarios, followed by developing Hidden Markov Model (HMM) and Partially Observable Monte-Carlo Planning (POMCP) algorithms for modeling the security attacks. Also, potential mitigation strategies are suggested for the identified threats.