计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (19): 1-13.DOI: 10.3778/j.issn.1002-8331.2204-0315

• 热点与综述 • 上一篇    下一篇

基于深度学习的属性图异常检测综述

张伊扬,钱育蓉,陶文彬,冷洪勇,李自臣,马梦楠   

  1. 1.新疆大学 软件学院,乌鲁木齐 830046
    2.新疆大学 新疆维吾尔自治区信号检测与处理重点实验室,乌鲁木齐 830046
    3.北京理工大学 计算机学院,北京 100081
    4.广东水利电力职业技术学院 大数据与人工智能学院,广州 510635
  • 出版日期:2022-10-01 发布日期:2022-10-01

Survey of Attribute Graph Anomaly Detection Based on Deep Learning

ZHANG Yiyang, QIAN Yurong, TAO Wenbin, LENG Hongyong, LI Zichen, MA Mengnan   

  1. 1.Software College, Xinjiang University, Urumqi 830046, China
    2.Key Laboratory of Signal Detection and Processing in Xinjiang Uygur Autonomous Region, Xinjiang University, Urumqi 830046, China
    3.School of Computer Science, Beijing Institute of Technology, Beijing 100081, China
    4.College of Big Data and Artificial Intelligence, Guangdong Polytechnic of Water Resources and Electric Engineering, Guangzhou 510635, China
  • Online:2022-10-01 Published:2022-10-01

摘要: 异常检测一直以来都是数据挖掘领域的研究热点之一,其任务是在海量数据中识别罕见的观测对象。随着图数据挖掘的发展,属性图异常检测在各个领域广受关注。然而,属性图因其复杂的拓扑结构和丰富的属性信息成为异常检测一大难点。深度学习方法在捕捉属性图复杂的信息中展现出优越性能,已被证实是解决属性图异常检测问题非常有效的方法。对普通图异常检测和属性图异常检测以及表示学习相关方法进行简要概述;其次从静态属性图和动态属性图两方面对最新深度学习异常检测方法进行介绍与分类;对常见数据集上的实验结果进行了对比、分析;对属性图异常检测的应用场景、存在的问题以及面临的挑战进行讨论,展望了未来的研究方向。

关键词: 异常检测, 属性图, 图数据挖掘, 深度学习

Abstract: Anomaly detection has always been one of the research hotspots in the field of data mining, and its task is to identify rare observations in massive data. With the development of graph data mining, attribute graph anomaly detection has received wide attention in various areas. However, attribute graphs have become a difficult problem in anomaly detection due to their complex topology and rich attribute information. Deep learning methods have shown superior performance in capturing complex information in attribute graphs, and have been proven to be very effective methods for solving the problem of anomaly detection in attribute graphs. Firstly, a brief overview of common graph anomaly detection, attribute graph anomaly detection and representation learning related methods are given. Secondly, the latest deep learning anomaly detection methods are introduced and classified as static and dynamic attribute graphs. Finally, the application scenarios, existing problems and challenges of attribute graph anomaly detection are discussed, and the future research directions are prospected.

Key words: anomaly detection, attribute graph, graph data mining, deep learning