Quickest Attack Detection in Smart Grid Based on Sequential Monte Carlo Filtering

TL;DR

This paper introduces a novel method combining sequential Monte Carlo filtering with CUSUM testing for rapid cyber attack detection in smart grids, improving detection speed and efficiency while reducing measurement communication.

Contribution

It proposes a joint state estimation and attack detection framework using SMC filtering and adaptive sampling, enhancing detection speed and reducing measurement overhead.

Findings

High detection accuracy for various attack types

Reduced measurement and communication rates

Fastest attack detection achieved

Abstract

Quick and accurate detection of cyber attack is key to the normal operation of the smart grid system. In this paper, a joint state estimation and sequential attack detection method for a given bus with grid frequency drift is proposed that utilizes the commonly monitored output voltage. In particular, based on a non-linear state-space model derived from the three-phase sinusoidal voltage equations, we employ the sequential Monte Carlo (SMC) filtering to estimate the system state. The output of the SMC filter is fed into a CUSUM test to detect the attack in a fastest way. Moreover, an adaptive sampling strategy is proposed to reduce the rate of taking measurements and communicating with the controller. Extensive simulation results demonstrate that the proposed method achieves high adaptivity and efficient detection of various types of attacks in power systems.