Detection of Faults and Attacks Including False Data Injection Attack in Smart Grid Using Kalman Filter

TLDR

Smart‑grid systems are vulnerable to attacks exploiting their communication infrastructure, including DoS, random, and data‑injection attacks, and the χ² detector has been proven effective when combined with a Kalman filter for monitoring dependent variables. This study develops a mathematical model and a robust security framework to detect smart‑grid faults and attacks, particularly addressing false data‑injection attacks. The framework employs a Kalman filter to estimate state variables, then uses the χ² detector or a proposed Euclidean detector on the estimates and system readings. The χ² detector identifies DoS and random attacks but fails to detect false data‑injection attacks, whereas the Euclidean detector successfully detects such sophisticated injection attacks.

Abstract

By exploiting the communication infrastructure among the sensors, actuators, and control systems, attackers may compromise the security of smart-grid systems, with techniques such as denial-of-service (DoS) attack, random attack, and data-injection attack. In this paper, we present a mathematical model of the system to study these pitfalls and propose a robust security framework for the smart grid. Our framework adopts the Kalman filter to estimate the variables of a wide range of state processes in the model. The estimates from the Kalman filter and the system readings are then fed into the χ ² -detector or the proposed Euclidean detector. The χ ² -detector is a proven effective exploratory method used with the Kalman filter for the measurement of the relationship between dependent variables and a series of predictor variables. The χ ² -detector can detect system faults/attacks, such as DoS attack, short-term, and long-term random attacks. However, the studies show that the χ ² -detector is unable to detect the statistically derived false data-injection attack. To overcome this limitation, we prove that the Euclidean detector can effectively detect such a sophisticated injection attack.