ABSTRACT To address the characteristics of wide distribution, limited resources, and complex environments at the edge nodes of power terminal embedded components, as well as the inadequacy of traditional attack tracing technologies in such scenarios, this paper designs a graph model‐based pilot attack tracing technology for edge nodes of power terminal embedded components. The method dynamically adjusts node trust using an anomaly attenuation factor and employs a random walk‐based graph embedding algorithm to convert attack behavior graphs into low‐dimensional features. Information entropy and mutual information quantify inter‐component information flow to support attack path visualization. Experiments show the method achieves a pilot attack tracing probability over 0.1, with detection probability rising from 0.963 to 0.999 across iterations, anti‐attack capability consistently above 0.97, and packet loss below 2.0%.
Xiao et al. (2026) studied this question.