Community-based anomaly detection using spectral graph filtering

Several applications have a community structure where the nodes of same share similar attributes. Anomaly or outlier detection in networks is relevant and widely studied research topic with various domains. Despite significant amount anomaly frameworks, there dearth on literature methods that consider both attributed graphs networks. This paper proposes community-based algorithm using spectral graph-based filter includes network into Laplacian matrix adopted as basis for Fourier transform. In addition, choice cutoff frequency considers number communities found. computational experiments, proposed strategy, called SpecF, showed an outstanding performance successfully identifying even discrete anomalies. SpecF better than baseline disregarding structure, especially higher overlapping. Additionally, we present case study to validate method dissemination COVID-19 different districts São José dos Campos, Brazil.