Towards Fair Graph Anomaly Detection: Problem, New Datasets, and Evaluation

arXiv:2402.15988v1 Announce Type: cross
Abstract: The Fair Graph Anomaly Detection (FairGAD) problem aims to accurately detect anomalous nodes in an input graph while ensuring fairness and avoiding biased predictions against individuals from sensitive subgroups such as gender or political leanings. Fairness in graphs is particularly crucial in anomaly detection areas such as misinformation detection in search/ranking systems, where decision outcomes can significantly affect individuals. However, the current literature does not comprehensively discuss this problem, nor does it provide realistic datasets …

abstract anomaly anomaly detection arxiv cs.lg cs.si datasets detection evaluation fair fairness gender graph graphs nodes political predictions subgroups type