A Comprehensive Analysis of GSP-Based Attack Detection Techniques Sensitivity to Topology Noise in Power Systems _*

Accurate graph representation of power systems is essential for many advanced graph-aware data analytics methods that support their critical functions. Graph Signal Processing (GSP) offers powerful tools for analyzing power system measurements and capturing interactions among these measurements via a graph. However, these graphs, like any data, are susceptible to noise. This study conducts a comprehensive sensitivity analysis of three prominent GSP-based false data injection attack (FDIA) detection techniques in response to topology noise. Topology noise is modeled as the addition or removal of a single line in the system's graph. The accuracy of these detection techniques is evaluated across various scenarios to understand the impact of the spatial distribution of topology noise on their performance. The results demonstrate that all the techniques are sensitive to topology noise, with some showing greater robustness depending on the spatial distribution of the noise.