PI: Bruno Sinopoli, Washington University St. Louis
Data integrity attacks forge power measurements or control commands in the power grid and can lead to wrong power state estimation and wrong control actions, which could be catastrophic. This project aims to detect data integrity attacks and perform robust estimation of the power system state under such attacks. Our first goal is to develop a robust passive detection filter, which cannot only detect attacks on the grid, but can also perform robust estimation in the presence of faulty or malicious behavior, thus enabling operators to perform adequate feedback control, even in the presence of attacks.
Second, we propose physical watermarking as a tool to detect replay attacks. In physical watermarking, the defender adds a secret random noisy input or watermark on top of the optimal input to perform detection. Under normal operation, one expects to see the watermark in the sensor outputs, since PMU measurements are physically correlated to control inputs through system dynamics. However, under replay attack, the watermark is no longer present. Detection algorithms will be designed to validate the presence of the watermark in the outputs and thus verify the health of the subsystem. The proposed algorithms can be implemented in software, allowing for easy plug and play.