Cyber attacks scenarios on the measurement function of power state estimation

Cyber security provided by robust power systems state estimation (SE) methods is evaluated. Tools and estimators developed in robust statistics theory are considered. The least trimmed squares (LTS) based diagnostic is studied. The impact of cyber attacks on the Jacobian matrix or observation function, which generate coordinated outliers known as leverage points, is assessed. Two scenarios of attacks are proposed where the first scenario generates a masked attack resulting in a contaminated state uncontrolled by the intruder. The second scenario leads to a stealthy attack with an estimated targeted state fixed by the same intruder. Intervals for the necessary number of attacks for each scenario and their positions are shown. Theoretical derivations based on a projection framework highlights the conditions that minimize detection with robust SE approaches developed from the regression model assumption. More specifically, affine equivariant robust estimators present a weakness towards such intrusions. Simulations on IEEE power system test beds illustrate the behavior of the robust LTS with decomposition and the popular detection methods analyzing the weighted least squares (WLS) residuals when subject to both scenarios' attacks.