Multi-dimensional Adaptive Mix-hop Contextual Learning Framework for Universal Graph Anomaly Detection

Graph Anomaly Detection focuses on identifying instances that deviate from normal patterns in graph-structured data. Although substantial progress has been made in this field, current approaches are constrained by the "one-dataset-onemodel" paradigm, exhibiting limited generalization across graphs with heterogeneous feature spaces, poor adaptability in few-shot scenarios, and inefficient cross-domain deployment. To overcome these limitations, we propose SAARCS, a universal graph anomaly detection framework capable of performing anomaly detection across diverse graph datasets without requiring any target data training. SAARCS aligns feature dimensions through composite spatial smoothness, learns graph embeddings via an adaptive-hop attention encoder, and predicts node abnormality using only a small set of normal context nodes. Extensive experiments on eight realworld datasets demonstrate that our approach achieves superior performance compared to state-of-the-art baselines.