Alleviating Structural Distribution Shift in Graph Anomaly Detection

Abstract

Graph anomaly detection (GAD) is a challenging binary classification problem due to its different structural distribution between anomalies and normal nodes -- abnormal nodes are a minority, therefore holding high heterophily and low homophily compared to normal nodes. Furthermore, due to various time factors and the annotation preferences of human experts, the heterophily and homophily can change across training and testing data, which is called structural distribution shift (SDS) in this paper. The mainstream methods are built on graph neural networks (GNNs), benefiting the classification of normals from aggregating homophilous neighbors, yet ignoring the SDS issue for anomalies and suffering from poor generalization. This work solves the problem from a feature view. We observe that the degree of SDS varies between anomalies and normal nodes. Hence to address the issue, the key lies in resisting high heterophily for anomalies meanwhile benefiting the learning of normals from homophily. We tease out the anomaly features on which we constrain to mitigate the effect of heterophilous neighbors and make them invariant. We term our proposed framework as Graph Decomposition Network (GDN). Extensive experiments are conducted on two benchmark datasets, and the proposed framework achieves a remarkable performance boost in GAD, especially in an SDS environment where anomalies have largely different structural distribution across training and testing environments. Codes are open-sourced in https://github.com/blacksingular/wsdm_GDN.

arXiv commentAccepted to WSDM 2023

References in corpus (14)

• Disentangling by Factorising
• Enhancing Graph Neural Network-based Fraud Detectors against Camouflaged Fraudsters
• A Semi-supervised Graph Attentive Network for Financial Fraud Detection
• Towards Out-Of-Distribution Generalization: A Survey
• From Amateurs to Connoisseurs: Modeling the Evolution of User Expertise through Online Reviews
• Unifying Graph Convolutional Neural Networks and Label Propagation
• Graph-based Anomaly Detection and Description: A Survey
• Causal Understanding of Fake News Dissemination on Social Media
• GraphLIME: Local Interpretable Model Explanations for Graph Neural Networks
• Rethinking Graph Neural Networks for Anomaly Detection
• Instance-Dependent PU Learning by Bayesian Optimal Relabeling
• Subgroup Generalization and Fairness of Graph Neural Networks
• Semi-supervised Embedding in Attributed Networks with Outliers
• When Does A Spectral Graph Neural Network Fail in Node Classification?

Cited by in corpus (6)

• Causal invariant geographic network representations with feature and structural distribution shifts
• Reconciling Attribute and Structural Anomalies for Improved Graph Anomaly Detection
• Lower-Left Partial AUC: An Effective and Efficient Optimization Metric for Recommendation
• Polynomial Selection in Spectral Graph Neural Networks: An Error-Sum of Function Slices Approach
• Partition-wise Graph Filtering: A Unified Perspective Through the Lens of Graph Coarsening
• Towards Understanding Link Predictor Generalizability Under Distribution Shifts