Counterfactual Data Augmentation with Denoising Diffusion for Graph Anomaly Detection

TL;DR

This paper introduces CAGAD, an unsupervised graph anomaly detection method that uses counterfactual data augmentation with a diffusion model to improve the distinguishability of anomalous node representations.

Contribution

It proposes a novel counterfactual data augmentation approach for GNNs, utilizing a diffusion model to generate more distinguishable anomaly representations.

Findings

CAGAD outperforms strong baselines in four datasets.

Average improvements of 2.35% in F1, 2.53% in AUC-ROC, and 2.79% in AUC-PR.

Uses a graph pointer neural network for anomaly detection.

Abstract

A critical aspect of Graph Neural Networks (GNNs) is to enhance the node representations by aggregating node neighborhood information. However, when detecting anomalies, the representations of abnormal nodes are prone to be averaged by normal neighbors, making the learned anomaly representations less distinguishable. To tackle this issue, we propose CAGAD -- an unsupervised Counterfactual data Augmentation method for Graph Anomaly Detection -- which introduces a graph pointer neural network as the heterophilic node detector to identify potential anomalies whose neighborhoods are normal-node-dominant. For each identified potential anomaly, we design a graph-specific diffusion model to translate a part of its neighbors, which are probably normal, into anomalous ones. At last, we involve these translated neighbors in GNN neighborhood aggregation to produce counterfactual representations of anomalies. Through aggregating the translated anomalous neighbors, counterfactual representations become more distinguishable and further advocate detection performance. The experimental results on four datasets demonstrate that CAGAD significantly outperforms strong baselines, with an average improvement of 2.35% on F1, 2.53% on AUC-ROC, and 2.79% on AUC-PR.