Detecting Anomalies by Data Aggregation in the Power Grid

The August 2003 Blackout event showed that the PowerGrid is vulnerable to the cyber attacks. The event also showed the need for automatic detection of abnormal behaviors in the PowerGrid. In this paper, we propose a solution for the problem of detecting anomaly behaviors such as reporting incorrect status of the devices of the PowerGrid. First, we apply the non-parametric Cumulative Sum algorithm for quickest detection of the value-changing problem. Then, to improve the speed of detection, we explore the use of two data aggregation schemes: average-aggregation and quantize-aggregation. Our analytical and simulation results show that end-to-end detection delay should include not only the semantic computation of the data but also the underlying communication networks. Finally, we show that it is possible to improve the end-to-end detection delay by data aggregation.