基于超球面对偶学习的双通道图异常检测方法  

Anomaly Detection with Dual-Channel Heterogeneous Graph NeuralNetwork Based on Hypersphere Dual Learning

作　　者：李青[1] 钟将[2] 倪航 LI Qing;ZHONG Jiang;NI Hang(School of Computer Science,Northwestern Polytechnical University,Xi’an,Shaanxi 710072,China;College of Computer Science,Chongqing University,Chongqing 400044,China)

机构地区：[1]西北工业大学计算机学院,陕西西安710072 [2]重庆大学计算机学院,重庆400044

出　　处：《电子学报》2024年第7期2212-2218,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.62102316,No.62171382);中央高校基本科研业务费资助项目(No.G2021KY05114);“十四五”共用信息系统装备预先研究项目(No.315197202);航空科学基金(No.20200051053002)~~。

摘　　要：图异常检测作为一项重要的数据挖掘任务,专注于识别与大多数节点显著偏离的异常节点.随着无监督图神经网络技术的进步,现已开发出了基于密度估计、对抗生成网络等多种高效识别图数据中潜在异常的方法 .然而,这些方法更注重无监督图异常检测生成高质量的表征,而往往忽略了图异常的特性.因此,本文提出了一个双通道异构图异常检测模型(Dual-channel Heterogeneous Graph Anomaly Detection,HD-GAD).其模型基础架构包括双通道的图神经网络:全局子结构感知的图神经网络和局部子结构感知的图神经网络,用于图异常检测捕获全局和局部子结构属性.同时,基于对偶推断引入了多超球体学习目标(Multi-Hypersphere Learning,MHL),从宏观和介观超球体角度,分别测量在整个图/社区结构中偏离的异常节点. HD-GAD模型利用相似度函数EmbSim优化训练目标,以缓解多超球面学习中的模型坍问题.最后,在五种不同的数据集上进行了全面的实验.其AUC(Area Under Curve)值在大多数情况下均超过了0.9,达到了行业领先水平,进一步证明了HD-GAD模型在图异常检测任务上的高效性与性能优势.Graph anomaly detection,as a crucial data mining task,focuses on identifying anomalous nodes that signif⁃icantly deviate from the majority of the nodes.With the advancement of unsupervised graph neural network techniques,var⁃ious efficient methods have been developed to detect potential anomalies in graph data,including those based on density es⁃timation and generative adversarial networks.However,these methods often focus on generating high-quality representa⁃tions for unsupervised graph anomaly detection and tend to overlook the characteristics of graph anomalies.Consequently,this paper proposes a dual-channel heterogeneous graph anomaly detection model(HD-GAD).Its architecture includes two graph neural networks,i.e.a global substructure-aware GNN(Graph Neural Network)and a local substructure-aware GNN,designed to capture global and local substructural properties for graph anomaly detection.Additionally,the model introduc⁃es a multi-hypersphere learning(MHL)objective based on dual inference,which measures anomalies deviating from the overall graph/community structure from macro and meso hypersphere perspectives.The HD-GAD model utilizes the simi⁃larity function EmbSim to optimize the training objective,mitigating model collapse issues in multi-hypersphere learning.Comprehensive experiments conducted on five different datasets demonstrated that the AUC(Area Under Curve)values ex⁃ceeded 0.9 in most cases,achieving industry-leading levels and further proving the HD-GAD model's efficiency and perfor⁃mance advantages in graph anomaly detection tasks.

关 键 词：图异常检测 图神经网络 超球面学习 双通道图神经网络 无监督学习 对偶学习 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]