Semi-supervised False Data Injection Attacks Detection in Smart Grid

Abstract

AbstractFalse data injection attacks (FDIAs) detection in smart grid, requires adequate labeled training samples to train a detection model. Due to the strong subjectivity, relying on expert knowledge and time-consuming nature of power system sample annotation, this task is intrinsically a small sample learning problem. In this paper, we propose a novel semi-supervised detection algorithm for FDIAs detection. The semi-supervised label propagation algorithm can dynamically propagate the label from labeled samples to unlabeled samples, automatically assign class labels to the unlabeled samples dataset, and enlarge the labeled samples dataset. Jointly use a small number of manually labeled samples dataset and a large number of auto-labeled samples dataset to construct a classifier via semi-supervised learning. Comparing the proposed algorithm with supervised learning algorithms, the results suggest that, with the scheme of semi-supervised learning from large unlabeled dataset, the proposed algorithm can significantly improve the accuracy of false data injection attacks detection.KeywordsFalse data injection attacksSemi-supervised learningLabel propagationSmall sample learning