Deep Adversary based Stealthy False Data Injection Attacks against AC state estimation

Conventional power systems are increasingly getting converted into smart cyber-physical systems with the incorporation of communication and information technologies. These changes are being made to ensure efficient control and reliability of operations under large-scale penetration of renewable energy-based generation and flexible loads like Electric Vehicles. However, these transformations are also responsible for making the power system increasingly susceptible to cyber threats like False Data Injection Attacks. In this paper, we propose a novel stealthy false data injection attack model by leveraging the concepts of adversarial attacks using just historical data collected from the power system. The power system state estimator is first approximated with a neural network which is then used to generate adversarial samples that are used for stealthy attack injection. Simulation results on the IEEE 118 bus test system show that the attack vectors generated using this model can introduce significant deviations in the estimated state values without triggering the conventional bad data detection safeguards.