Cyber intrusion detection through spatio-temporal correlation in optimal power flow problem

The proposed approach formulates a density based value to judge each new incoming measurement as spatio-temporal outliers among all the measurements obtained in the same power network. These spatio-temporal outliers are then detected via a density-based local outlier detection algorithm. The proposed approach takes advantage of the spatio-temporal correlations among multiple time instants of synchrophasor units, and detects data integrity attacks. It is purely data-driven algorithm without involving topology of the power grid, which also avoids the effect of inaccurate system information on the detection results. The algorithm is able to detect data integrity attacks under both normal and abnormal operating conditions and also suitable for real-time application. The effectiveness of the proposed approach is verified through 14 bus system case study using simulated data set.