High-fidelity Intrusion Detection Datasets for Smart Grid Cybersecurity Research

Intrusion Detection Systems (IDSes) are key defense mechanisms for securing smart grids against cyberattacks. They require realistic datasets to develop accurate models for detecting network anomalies. However, acquiring realistic datasets is challenging due to the need for expert knowledge to accurately label attack data, the sensitive nature of the information, and safety issues related to attacking the actual power systems. Consequently, there is a lack of high-fidelity datasets for testing and validating the efficacy of IDSes. While synthetic datasets provide a good workaround, they are often unrealistic and fail to capture the physical dynamics of power systems under cyberattacks. To address this gap, we leverage the Electric Power and Intelligent Control (EPIC) testbed, a hardware-based smart grid security testbed, to simulate false data injection attacks (FDIA) and time delay attacks (TDA) on two critical power grid operations, namely generator synchronization and reverse power prevention. Our goal is to generate representative datasets that accurately model those operations under normal and attack conditions. By making these datasets publicly available, we enable the research community to develop more effective IDS solutions to enhance the security of smart grids.