A Zero-shot Generalized Graph Anomaly Detection Framework via Node Reconstruction

The article introduces AlignGAD, a zero-shot generalized graph anomaly detection framework designed to identify abnormal nodes in unseen target graphs without relying on dataset-specific features. The framework consists of a Global Unification Module for feature alignment, a Clustering Module for capturing group-level anomalies, and a Node Discrepancy Scoring Module for measuring reconstruction discrepancies. Experimental results on various real-world datasets show that AlignGAD effectively generalizes across different domains, making it valuable for practitioners working with heterogeneous graph data.