Injection Attacks and Detection Strategy in Front-End Vehicle-to-Grid Communication

Public electric vehicle (EV) charging stations provide accessible charging options and play a vital role in addressing range anxiety and facilitating long-distance travel. However, the wide adoption of public charging stations poses serious security risks. This paper demonstrates for the first time an injection attack on the front-end vehicle-to-grid (V2G) communication based on the ISO 15118 protocol. Specifically, we developed a testbed that integrates V2Gdecoder, Parasite6, Open vSwitch, and MiniV2G to emulate traffic injections between the supply equipment communication controller (SECC) at a charging station and the EV's communication controller (EVCC). We showed that a malicious EV owner or infected supply equipment can inject harmful packets into the other side. This injection attack can modify the V2G messages to include runtime and denial-of-service instances, remote code executions, and other malware. To design a defense mechanism, we study the development of a machine learning-based system that can detect such injection attacks. We created a dataset of three cyber features that represent benign and malicious traffic between the SECC and EVCC. Then, we developed shallow and deep-learning supervised models that can detect injection attacks on front-end V2G traffic with detection rates up to 95% and false alarm rates down to 7%. Our experimental results highlight the potential of machine learning-based intrusion detection systems to effectively detect injection attacks on front-end V2G communications.