Data driven approach to attack detection in a cyber-physical smart grid system

Smart grid infrastructure is an integration of advanced communication, sensing and computing techniques into the existing physical electrical grid. It is emerging as a critical cyber-physical system (CPS) infrastructure. CPS comes across as systems with a tight integration between the physical and the cyber layers in addition to the communication network. This exposes it to a risk of mis-operation when subjected to false or malicious data injection from intruders or adversaries. In order to address this issue, this paper proposes a data analytics based cyber defense strategy of a substation automation system (SAS) by developing an intelligent module which uses Support vector machine (SVM) along with Principal component analysis (PCA). We propose a two stage approach for detecting the false data injection by an intruder into the system. First stage reduces the high dimensional dataset to a lower dimension using PCA and the second stage uses SVM on the reduced dataset for detecting the malicious attempts of introducing false data into the system. The proposed module uses the measurement values received from the Phasor Measurement Units (PMUs). Upon receiving a fault signal, the protection system initiates checks on incoming data patterns and compares it with the behavior of system dynamics. Using data analytics tools the fault condition signaling is classified as actual (real) fault or false (fake) fault with malicious intentions. The proposed methodology is demonstrated on three bus power network.