Autoencoder-based Approach to Detect Stealth Cyberattacks in Battery Energy Storage Systems

This paper addresses the detection of anomalies caused by stealth data injection attacks in battery energy storage systems (BESS). The proposed solution relies on autoencoders (AE) to learn the normal behavior of the system and then verify whether an anomaly caused by an attack is occurring. The proposed approach is assessed using data from a real BESS. The results indicate that the AE-based approach is a promising solution to detect anomalies caused by data injection attacks, even if it uses stealth techniques, thereby enhancing the overall operational efficiency and reliability of the BESS.