A Zero-shot Generalized Graph Anomaly Detection Framework via Node Reconstruction
Researchers have introduced AlignGAD, a novel framework designed for zero-shot generalized graph anomaly detection. This approach aims to identify abnormal nodes in new, unseen graphs by overcoming the limitations of existing methods that are often domain-specific. AlignGAD utilizes a Global Unification Module to align node features and graph signals, a Clustering Module to identify group-level abnormal patterns, and a Node Discrepancy Scoring Module to aggregate anomaly evidence. Experimental results on various real-world datasets indicate that AlignGAD is effective in zero-shot graph anomaly detection scenarios. AI