Subtractive Aggregation for Attributed Network Anomaly Detection (AAGNN)

1.Introduction

This repository contains code for paper "Subtractive Aggregation for Attributed Network Anomaly Detection" (CIKM'21). This repository contains code for paper "Subtractive Aggregation for Attributed Network Anomaly Detection" (CIKM'21).

2. Usage

Requirements:

• pytorch==1.10.0
• scikit-learn
• networkx
• scipy
• pandas

Datasets:

Users can create datasets with injected anomalies by themselves. For details (e.g., code), users can refer to this paper.

Examples:

• python main.py --dataset=BlogCatalog_anomaly --model=Atten_Aggregate --seed=1
• python main.py --dataset=BlogCatalog_anomaly --seed=1

Evaluation:

This code performs evaluation on the test set (e.g., 50% data). When comparing with unsupervised methods, users should keep the same data volume of the test set (i.e., either on all the nodes or the test set).

3. Citation

Please kindly cite the paper if you use the code or any resources in this repo:

@inproceedings{zhou2021subtractive,
  title={Subtractive aggregation for attributed network anomaly detection},
  author={Zhou, Shuang and Tan, Qiaoyu and Xu, Zhiming and Huang, Xiao and Chung, Fu-Lai},
  booktitle={Proceedings of the 30th ACM International Conference on Information \& Knowledge Management},
  pages={3672--3676},
  year={2021}
}