ANN-Based Stealth Attack to Battery Energy Storage Systems by Using a Low-Cost Device

The application of the Internet of Things paradigm to the power and energy sector is opening the way to advanced monitoring and control capabilities over Distributed Energy Resources (DERs). In particular, the ability to use heterogeneous communication infrastructures is a key feature to enable the integration of DERs in modern distribution networks, but also expose them to cyber threats. Among the variety of threats warned by the literature, stealth attacks are one of the most relevant to be studied and faced. In this work, we present a Man-in-the-Middle (MitM) attack to a distributed Battery Energy Storage System (BESS) based on an Artificial Neural Network (ANN), launched from a low-cost device. The attack aims to stealthily harm the BESS by, e.g., charging and discharging the batteries to reduce their lifetime. During the attack, the ANN-based model is used to mimic the normal behavior of the BESS by hacking the communication of its State of Charge (SOC). The model is trained using the real SOC data as a function of power request commands sent to the BESS. The results show that the attack is able to manipulate the BESS and mimic its normal behavior with a mean absolute error of 1.19%. The evaluation on the computational resources indicates that the proposed attack can be deployed in a low-cost intelligent electronic device (e.g., a Raspberry Pi).

Automated summary

This paper presents a Man-in-the-Middle attack on a distributed Battery Energy Storage System (BESS) based on an Artificial Neural Network (ANN), which manipulates the system behavior by altering the State of Charge (SOC) communication. The results show that the attack is able to mimic the normal behavior of the BESS with a small error.