Generalized FDIA Detection in Power Dependent Electrified Transportation Systems

The increasing popularity of electric vehicles (EVs) has strengthened the integration between power and transportation networks. Amidst the rising trend of interconnecting power and transportation networks, there has been a surge in attacks targeting power systems, specifically those capable of disrupting charging services. Existing studies on identifying false data injection attacks (FDIAs) are insufficient in safeguarding coupled power-dependent electrified transportation networks for several reasons: (a) they are primarily designed for power networks only, (b) they does not consider the influence of cyber attacks on charging satisfaction rates, and (c) they lack adaptability/do not generalize well to various topological configurations. To overcome these shortcomings, this paper introduces a comprehensive FDIA detection scheme that takes into account the interconnected nature of power-transportation infrastructures, eventually improving the charging user satisfaction rates. Toward this objective, we develop a defense strategy based on a graph autoencoder (GAE) that extracts spatio-temporal features from the intertwined data, thereby providing increased resilience against FDIAs. Furthermore, our model undergoes training on diverse power system topologies and various attack scenarios, ensuring improved generalization capabilities. Simulations were conducted on two types of power systems: one with 2000 buses and another with 30 buses, featuring 360 and 35 charging stations (CSs) respectively. When subjected to unseen data, our model achieved an impressive 98.3% detection rate, marking a significant enhancement of 15% to 30% compared to benchmark strategies. This highlights the efficacy of our proposed method in adequately tackling the challenges associated with detecting FDIAs on interconnected power and transportation networks.