TY - GEN
T1 - Detection of False Data Injection Attacks in Unobservable Power Systems by Laplacian Regularization
AU - Dabush, Lital
AU - Routtenberg, Tirza
N1 - Funding Information:
This research was supported by the Israel Ministry of Infrastructure, Energy, and Water Resources. L. Dabush is a Fellow of the Ariane de Rothschild Women’s Doctoral Program.
Publisher Copyright:
© 2022 IEEE.
PY - 2022/1/1
Y1 - 2022/1/1
N2 - The modern electrical grid is a complex cyber-physical system, and thus is vulnerable to measurement losses and attacks. In this paper, we consider the problem of detecting false data injection (FDI) attacks and bad data in unobservable power systems. Classical bad-data detection methods usually assume observable systems and cannot detect stealth FDI attacks. We use the smoothness property of the system states (voltages) w.r.t. the admittance matrix, which is also the Laplacian of the graph representation of the grid. First, we present the Laplacian-based regularized state estimator, which does not require full observability of the network. Then, we derive the Laplacian-regularized generalized likelihood ratio test (LR-GLRT). We show that the LR-GLRT has a component of a soft high-pass graph filter applied to the state estimator. Numerical results on the IEEE 118-bus system demonstrate that the LR-GLRT outperforms other detection approaches and is robust to missing data.
AB - The modern electrical grid is a complex cyber-physical system, and thus is vulnerable to measurement losses and attacks. In this paper, we consider the problem of detecting false data injection (FDI) attacks and bad data in unobservable power systems. Classical bad-data detection methods usually assume observable systems and cannot detect stealth FDI attacks. We use the smoothness property of the system states (voltages) w.r.t. the admittance matrix, which is also the Laplacian of the graph representation of the grid. First, we present the Laplacian-based regularized state estimator, which does not require full observability of the network. Then, we derive the Laplacian-regularized generalized likelihood ratio test (LR-GLRT). We show that the LR-GLRT has a component of a soft high-pass graph filter applied to the state estimator. Numerical results on the IEEE 118-bus system demonstrate that the LR-GLRT outperforms other detection approaches and is robust to missing data.
KW - Graph signal processing (GSP)
KW - attack detection
KW - bad-data detection
KW - false data injection (FDI) attacks
KW - power system state estimation (PSSE)
UR - http://www.scopus.com/inward/record.url?scp=85135379829&partnerID=8YFLogxK
U2 - 10.1109/SAM53842.2022.9827810
DO - 10.1109/SAM53842.2022.9827810
M3 - Conference contribution
AN - SCOPUS:85135379829
T3 - Proceedings of the IEEE Sensor Array and Multichannel Signal Processing Workshop
SP - 415
EP - 419
BT - 2022 IEEE 12th Sensor Array and Multichannel Signal Processing Workshop, SAM 2022
PB - Institute of Electrical and Electronics Engineers
T2 - 12th IEEE Sensor Array and Multichannel Signal Processing Workshop, SAM 2022
Y2 - 20 June 2022 through 23 June 2022
ER -