- International Journal of Information Security Science
- Vol: 11 Issue: 4
- Performance Analysis of PCA Based Machine Learning Approaches on FDIA Detection
Performance Analysis of PCA Based Machine Learning Approaches on FDIA Detection
Authors : Kübra Bitirgen, Ümmühan Başaran Filik
Pages : 1-13
View : 10 | Download : 3
Publication Date : 2022-12-31
Article Type : Research
Abstract :Smart grid (SG) and its specific structures are widely taken notice of by many researchers studying power systems. This paper compares and analyzes the performance of five machine learning approaches combined with principal component analysis (PCA) to do the task of false data injection attack (FDIA) detection of an SG. For this purpose, PCA method combinations are presented and tested by using labeled data. Phasor measurement unit (PMU) data is a critical source of monitoring of progress and performance of an SG system. PMUs are perniciously influenced by FDIAs trying to manipulate the measurements without being noticed by the bad data detector (BDD) of the SG system. In one sense, the selected PMU data consisting of various features which play an important role in the control system of SG is used to analyze the characteristics of the SG system. The results show that FDIA detection is effectively accomplished. The efficiency of the proposed hybrid PCA-based various machine learning approaches is illustrated on a real measured PMU dataset. As empirical results show, Random Forest (RF) with PCA achieves the entire accuracy of 95% in FDIA detection.Keywords : False data injection attack (FDIA), Phasor measurement unit (PMU), Principle component analysis (PCA), Smart grid (SG)