Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Higher-order Structure Based Anomaly Detection on Attributed Networks

Xu Yuan, Na Zhou, Shuo Yu, Huafei Huang, Zhikui Chen, Feng Xia

TL;DR

This work tackles anomaly detection in attributed networks by leveraging higher-order structures (motifs) through GUIDE, a dual autoencoder framework that jointly reconstructs node attributes and higher-order structures. A graph node attention mechanism gates the learning of higher-order patterns, enabling effective detection of structural and attribute anomalies via reconstruction errors. GUIDE demonstrates superior ROC-AUC, PR-AUC, and Recall@K performance on five real-world datasets with injected anomalies, underscoring the value of incorporating motif-based representations. The approach offers a scalable, unsupervised solution that captures complex interactions among multiple entities, enhancing practical anomaly detection in networks with rich attribute and structural information.

Abstract

Anomaly detection (such as telecom fraud detection and medical image detection) has attracted the increasing attention of people. The complex interaction between multiple entities widely exists in the network, which can reflect specific human behavior patterns. Such patterns can be modeled by higher-order network structures, thus benefiting anomaly detection on attributed networks. However, due to the lack of an effective mechanism in most existing graph learning methods, these complex interaction patterns fail to be applied in detecting anomalies, hindering the progress of anomaly detection to some extent. In order to address the aforementioned issue, we present a higher-order structure based anomaly detection (GUIDE) method. We exploit attribute autoencoder and structure autoencoder to reconstruct node attributes and higher-order structures, respectively. Moreover, we design a graph attention layer to evaluate the significance of neighbors to nodes through their higher-order structure differences. Finally, we leverage node attribute and higher-order structure reconstruction errors to find anomalies. Extensive experiments on five real-world datasets (i.e., ACM, Citation, Cora, DBLP, and Pubmed) are implemented to verify the effectiveness of GUIDE. Experimental results in terms of ROC-AUC, PR-AUC, and Recall@K show that GUIDE significantly outperforms the state-of-art methods.

Higher-order Structure Based Anomaly Detection on Attributed Networks

TL;DR

This work tackles anomaly detection in attributed networks by leveraging higher-order structures (motifs) through GUIDE, a dual autoencoder framework that jointly reconstructs node attributes and higher-order structures. A graph node attention mechanism gates the learning of higher-order patterns, enabling effective detection of structural and attribute anomalies via reconstruction errors. GUIDE demonstrates superior ROC-AUC, PR-AUC, and Recall@K performance on five real-world datasets with injected anomalies, underscoring the value of incorporating motif-based representations. The approach offers a scalable, unsupervised solution that captures complex interactions among multiple entities, enhancing practical anomaly detection in networks with rich attribute and structural information.

Abstract

Anomaly detection (such as telecom fraud detection and medical image detection) has attracted the increasing attention of people. The complex interaction between multiple entities widely exists in the network, which can reflect specific human behavior patterns. Such patterns can be modeled by higher-order network structures, thus benefiting anomaly detection on attributed networks. However, due to the lack of an effective mechanism in most existing graph learning methods, these complex interaction patterns fail to be applied in detecting anomalies, hindering the progress of anomaly detection to some extent. In order to address the aforementioned issue, we present a higher-order structure based anomaly detection (GUIDE) method. We exploit attribute autoencoder and structure autoencoder to reconstruct node attributes and higher-order structures, respectively. Moreover, we design a graph attention layer to evaluate the significance of neighbors to nodes through their higher-order structure differences. Finally, we leverage node attribute and higher-order structure reconstruction errors to find anomalies. Extensive experiments on five real-world datasets (i.e., ACM, Citation, Cora, DBLP, and Pubmed) are implemented to verify the effectiveness of GUIDE. Experimental results in terms of ROC-AUC, PR-AUC, and Recall@K show that GUIDE significantly outperforms the state-of-art methods.
Paper Structure (16 sections, 14 equations, 8 figures, 4 tables, 1 algorithm)

This paper contains 16 sections, 14 equations, 8 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Network Motif
  5. Problem Definition
  6. The Proposed Model - GUIDE
  7. Attribute Autoencoder
  8. Structure Autoencoder
  9. Loss Function and Anomaly Detection
  10. Experiments
  11. Datasets
  12. Experimental Settings
  13. Experimental Results
  14. Ablation Experiment
  15. Parameter Analysis
  16. ...and 1 more sections

Figures (8)

  • Figure 1: The illustration about different types of anomalies on attributed networks.
  • Figure 2: A node motif degree calculation of M31 (For node e, it contains three M31 motifs, i.e., ebd, eac, ecd).
  • Figure 3: Network motifs used in this paper.
  • Figure 4: The framework of the proposed GUIDE.
  • Figure 5: ROC curves and AUC scores of all methods on five real-world datasets.
  • ...and 3 more figures

Theorems & Definitions (3)

  • Definition 1
  • Definition 2
  • Definition 3