Towards Multiple Missing Values-resistant Unsupervised Graph Anomaly Detection

TL;DR

This paper introduces M$^2$V-UGAD, a novel framework for unsupervised graph anomaly detection that is robust to missing node attributes and structure, using dual-pathway encoding and hard negative sampling.

Contribution

It proposes a dual-pathway encoder and a joint latent space regularization to effectively detect anomalies on incomplete graphs, reducing imputation bias and cross-view interference.

Findings

Outperforms existing methods on seven benchmarks.

Maintains high detection accuracy across various missing rates.

Effectively handles simultaneous missing attributes and edges.

Abstract

Unsupervised graph anomaly detection (GAD) has received increasing attention in recent years, which aims to identify data anomalous patterns utilizing only unlabeled node information from graph-structured data. However, prevailing unsupervised GAD methods typically presuppose complete node attributes and structure information, a condition hardly satisfied in real-world scenarios owing to privacy, collection errors or dynamic node arrivals. Existing standard imputation schemes risk "repairing" rare anomalous nodes so that they appear normal, thereby introducing imputation bias into the detection process. In addition, when both node attributes and edges are missing simultaneously, estimation errors in one view can contaminate the other, causing cross-view interference that further undermines the detection performance. To overcome these challenges, we propose M$^2$V-UGAD, a multiple missing values-resistant unsupervised GAD framework on incomplete graphs. Specifically, a dual-pathway encoder is first proposed to independently reconstruct missing node attributes and graph structure, thereby preventing errors in one view from propagating to the other. The two pathways are then fused and regularized in a joint latent space so that normals occupy a compact inner manifold while anomalies reside on an outer shell. Lastly, to mitigate imputation bias, we sample latent codes just outside the normal region and decode them into realistic node features and subgraphs, providing hard negative examples that sharpen the decision boundary. Experiments on seven public benchmarks demonstrate that M$^2$V-UGAD consistently outperforms existing unsupervised GAD methods across varying missing rates.