Cross-Domain Cybersecurity in Smart Grids: Fault-Tolerant Architecture and Anomaly Detection for V2G Infrastructure

Abstract

Decentralized Smart Grid architectures integrate Vehicle-to-Grid (V2G) infrastructure as active energy resources. Cybersecurity analyses of digital substations demonstrate vulnerabilities to False Data Injection (FDI) and time synchronization attacks. The V2G domain exhibits parallel exposure, lacking the deterministic security controls of isolated power grids. This paper maps cyber-physical attack vectors from operational technology (OT) networks onto the Open Charge Point Protocol (OCPP) and ISO 15118 standards. We propose a fault-tolerant architecture for V2G aggregators combining an edge-based Machine Learning Intrusion Detection System (IDS) and a multi-homed (4G/5G/Ethernet) Triple Modular Redundancy (TMR) mechanism for mitigating control and timing manipulation. The detection models are validated using empirical dataset acquired from a 22 kW bidirectional charging station (Infypower) and an electric vehicle (KIA EV3). Analysis of physical metrics—including True Power Factor, harmonic distortion, and reactive power transients—identifies load-altering attacks with an F1-score of 96.4%. The implemented TMR consensus algorithm isolates Byzantine faults within 500 ms, preventing adversarial destabilization of the local microgrid.