[1]王天昊,窦浩,王翔,等.融合图滤波与注意力机制的图异常检测方法[J].智能系统学报,2026,21(4):919-931.[doi:10.11992/tis.202510033]
WANG TianHao,DOU Hao,WANG Xiang,et al.Graph anomaly detection method integrating graph filtering and an attention mechanism[J].CAAI Transactions on Intelligent Systems,2026,21(4):919-931.[doi:10.11992/tis.202510033]
点击复制
WANG TianHao,DOU Hao,WANG Xiang,et al.Graph anomaly detection method integrating graph filtering and an attention mechanism[J].CAAI Transactions on Intelligent Systems,2026,21(4):919-931.[doi:10.11992/tis.202510033]
融合图滤波与注意力机制的图异常检测方法
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
21
期数:
2026年第4期
页码:
919-931
栏目:
学术论文—机器学习
出版日期:
2026-07-05
- Title:
- Graph anomaly detection method integrating graph filtering and an attention mechanism
- Keywords:
- graph data; graph neural networks; anomaly detection; heterophilic connection; feature inconsistency; graph filtering; attention mechanism; frequency domain information fusion
- 分类号:
- TP391
- 摘要:
- 为缓解图神经网络异常检测中普遍存在的异配连接与特征不一致问题，本文提出了一种结合图滤波与注意力机制的图异常检测方法(flexible high-low frequency and attention neural network, FHANN)。针对异配连接问题，设计了图滤波模块，可以自适应提取高频与低频信号并融合生成节点嵌入；对于特征不一致问题，从节点特征的相似度出发，引入图注意力模块用于学习节点嵌入；最后将两模块得到的节点嵌入进行融合并用于图异常检测任务。在3个真实数据集的实验结果表明，FHANN在AUC-ROC(area under the receiver operating characteristic curve)和AUC-PR(area under the precision-recall curve)两个评估指标上平均提高了1.17%和4.25%。本文研究成果可为社交网络分析、金融欺诈检测等复杂图结构数据的异常识别任务提供方法借鉴与技术参考。
- Abstract:
- To address the pervasive issues of heterophilic connections and feature inconsistency in graph neural network–based anomaly detection, this paper proposes the flexible high–low-frequency and attention neural network (FHANN), a graph anomaly detection method that integrates graph filtering and attention mechanisms. Specifically, to mitigate the effects of heterophilic connections, a graph filtering module is designed to adaptively capture and fuse high- and low-frequency signals for node representation learning. To address feature inconsistency, a graph attention module is introduced to learn the importance of node features and generate node representations based on feature similarity. The representations produced by the two modules are then fused for anomaly detection. Experiments on three real-world graph anomaly detection datasets demonstrate that FHANN achieves average improvements of 1.17% and 4.25% in area under the receiver operating characteristic curve(AUC-ROC) and area under the precision-recall curve(AUC-PR), respectively. These results verify the effectiveness of the proposed method and provide valuable insights for anomaly detection in complex graph-structured data, including social network analysis and financial fraud detection.
- 参考文献/References:
-
[1] 李忠, 靳小龙, 庄传志, 等. 面向图的异常检测研究综述[J]. 软件学报, 2021, 32(1): 167-193 LI Zhong, JIN Xiaolong, ZHUANG Chuanzhi, et al. A survey on graph-based anomaly detection research[J]. Journal of software, 2021, 32(1): 167-193
[2] 陈波冯, 李靖东, 卢兴见, 等. 基于深度学习的图异常检测技术综述[J]. 计算机研究与发展, 2021, 58(7): 1436-1455 CHEN Bofeng, LI Jingdong, LU Xingjian, et al. A survey on graph anomaly detection techniques based on deep learning[J]. Journal of computer research and development, 2021, 58(7): 1436-1455
[3] CUI Jiafu, YANG Siqi, YI Litai, et al. Recent advances in deep learning for protein-protein interaction: a review[J]. BioData mining, 2025, 18(1): 43
[4] KUMAR J S, ARCHANA B, KUMAR K M V S. Graph theory: modelling and analyzing complex system[J]. Metallurgical and materials engineering, 2025, 31(3): 70-77
[5] MOTIE S, RAAHEMI B. Financial fraud detection using graph neural networks: a systematic review[J]. Expert systems with applications, 2024, 240: 122156
[6] XUAN TUNG N, TUNG GIANG L, DUC SON B, et al. Graph neural networks for next-generation-IoT: recent advances and open challenges[J]. IEEE communications surveys & tutorials, 2026, 28: 2226-2262
[7] WANG Zhen, LAN Chao. Towards a hierarchical Bayesian model of multi-view anomaly detection[C]//Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence. Yokohama: International Joint Conferences on Artificial Intelligence Organization, 2020.
[8] PANG Guansong, SHEN Chunhua, CAO Longbing, et al. Deep learning for anomaly detection: a review[J]. ACM computing surveys, 2022, 54(2): 1-38
[9] WANG Daixin, LIN Jianbin, CUI Peng, et al. A semi-supervised graph attentive network for financial fraud detection[C]//2019 IEEE International Conference on Data Mining. Beijing: IEEE, 2019.
[10] 郭嘉琰, 李荣华, 张岩, 等. 基于图神经网络的动态网络异常检测算法[J]. 软件学报, 2020, 31(3): 748-762 GUO Jiayan, LI Ronghua, ZHANG Yan, et al. Dynamic network anomaly detection algorithm based on graph neural networks[J]. Journal of software, 2020, 31(3): 748-762
[11] XU Yiming, PENG Zhen, SHI Bin, et al. Revisiting graph contrastive learning on anomaly detection: a structural imbalance perspective[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2025, 39(12): 12972-12980
[12] LIU Ziqi, CHEN Chaochao, YANG Xinxing, et al. Heterogeneous graph neural networks for malicious account detection[C]//Proceedings of the 27th ACM International Conference on Information and Knowledge Management. Torino: ACM, 2018.
[13] DING Kaize, LI Jundong, BHANUSHALI R, et al. Deep anomaly detection on attributed networks[C]//Proceedings of the 2019 SIAM International Conference on Data Mining. Philadelphia: SIAM, 2019.
[14] CHEN Linghao, LI He, ZHANG Wanyuan, et al. AnomMAN: detect anomalies on multi-view attributed networks[J]. Information sciences, 2023, 628: 1-21
[15] 李贺, 彭以冲, 张万园, 等. 基于图卷积自编码器的多视图属性网络异常检测算法[J]. 中国科学: 信息科学, 2025, 55(2): 269-283 LI He, PENG Yichong, ZHANG Wanyuan, et al. Detect anomalies on multi-view attributed networks based on graph convolution autoencoder[J]. Scientia sinica informations, 2025, 55(2): 269-283
[16] 张铭泉, 周辉, 曹锦纲. 基于注意力机制的双BERT有向情感文本分类研究[J]. 智能系统学报, 2022, 17(6): 1220-1227 ZHANG Mingquan, ZHOU Hui, CAO Jingang. Dual BERT directed sentiment text classification based on attention mechanism[J]. CAAI transactions on intelligent systems, 2022, 17(6): 1220-1227
[17] 于润羽, 李雅文, 李昂. 融合领域特征的科技学术会议语义相似性计算方法[J]. 智能系统学报, 2022, 17(4): 737-743 YU Runyu, LI Yawen, LI Ang. Semantic similarity computing for scientific and technological conferences[J]. CAAI transactions on intelligent systems, 2022, 17(4): 737-743
[18] 马甜甜, 杨长春, 严鑫杰, 等. 融合知识图谱和轻量级图卷积网络推荐系统的研究[J]. 智能系统学报, 2022, 17(4): 721-727 MA Tiantian, YANG Changchun, YAN Xinjie, et al. Research on the fusion of knowledge graph and lightweight graph convolutional network recommendation system[J]. CAAI transactions on intelligent systems, 2022, 17(4): 721-727
[19] 陈容珊, 高淑萍, 齐小刚. 注意力机制和图卷积神经网络引导的谱聚类方法[J]. 智能系统学报, 2023, 18(5): 936-944 CHEN Rongshan, GAO Shuping, QI Xiaogang. A spectral clustering based on GCNs and attention mechanism[J]. CAAI transactions on intelligent systems, 2023, 18(5): 936-944
[20] LIU Yang, AO Xiang, QIN Zidi, et al. Pick and choose: a GNN-based imbalanced learning approach for fraud detection[C]//Proceedings of the Web Conference 2021. Ljubljana: ACM, 2021.
[21] HUANG Mengda, LIU Yang, AO Xiang, et al. AUC-oriented graph neural network for fraud detection[C]//Proceedings of the ACM Web Conference 2022. Online: ACM, 2022.
[22] 苏世玉, 于卿, 李婧, 等. 基于双重分类和重建的跨域图异常检测[J]. 计算机科学, 2025, 52(8): 375-385 SU Shiyu, YU Jiong, LI Shu, et al. Cross-domain graph anomaly detection via dual classification and reconstruction[J]. Computer science, 2025, 52(8): 375-385
[23] CHAI Ziwei, YOU Siqi, YANG Yang, et al. Can abnormality be detected by graph neural networks?[C]//Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence. Vienna: International Joint Conferences on Artificial Intelligence Organization, 2022.
[24] TANG Jianheng, LI Jiajin, GAO Ziqi, et al. Rethinking graph neural networks for anomaly detection[C]//Proceedings of the International Conference on Machine Learning. Hyderabad: IMLS, 2022.
[25] AKOGLU L, TONG Hanghang, KOUTRA D. Graph based anomaly detection and description: a survey[J]. Data mining and knowledge discovery, 2015, 29(3): 626-688
[26] KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[C]//Proceedings of the International Conference on Learning Representations. San Juan Capistrano: ICLR, 2016.
[27] VELICKOVIC P, CUCURULL G, CASANOVA A, et al. Graph attention networks[C]//Proceedings of the International Conference on Learning Representations. Vancouver: ICLR, 2018.
[28] HAMILTON W, YING Z, LESKOVEC J. Inductive representation learning on large graphs[C]//Proceedings of the Advances in Neural Information Processing Systems. Long Beach: NIPS Foundation, 2017.
[29] BRUNA J, ZREMBA W, SZLAM A, et al. Spectral networks and locally connected networks on graphs[C]// Proceedings of the International Conference on Learning Representations. Aspen: ICLR, 2014.
[30] DEFFERRARD M, BRESSON X, VANDERGHEYNST P. Convolutional neural networks on graphs with fast localized spectral filtering[C]//Proceedings of the Advances in Neural Information Processing Systems. Barcelona: NeurIPS Foundation, 2016.
[31] XU Bingbing, SHEN Huawei, CAO Qi, et al. Graph wavelet neural network[C]//Proceedings of the International Conference on Learning Representations. New Orleans: ICLR, 2019.
[32] LI Ao, QIN Zhou, LIU Runshi, et al. Spam review detection with graph convolutional networks[C]//Proceedings of the 28th ACM International Conference on Information and Knowledge Management. Beijing: ACM, 2019.
[33] GAO Yuan, WANG Xiang, HE Xiangnan, et al. Alleviating structural distribution shift in graph anomaly detection[C]//Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining. Singapore: ACM, 2023.
[34] MESGARAN M, BEN HAMZA A. Graph fairing convolutional networks for anomaly detection[J]. Pattern recognition, 2024, 145: 109960
[35] ROY A, SHU Juan, LI Jia, et al. GAD-NR: graph anomaly detection via neighborhood reconstruction[C]//Proceedings of the 17th ACM International Conference on Web Search and Data Mining. Merida: ACM, 2024.
[36] GAO Yuan, WANG Xiang, HE Xiangnan, et al. Addressing heterophily in graph anomaly detection: a perspective of graph spectrum[C]//Proceedings of the ACM Web Conference 2023. Austin: ACM, 2023.
[37] XU Fan, WANG Nan, WU Hao, et al. Revisiting graph-based fraud detection in sight of heterophily and spectrum[C]//Proceedings of the AAAI conference on artificial intelligence. Vancouver: AAAI, 2024.
[38] ZHENG Jianbo, YANG Chao, ZHANG Tairui, et al. Dynamic spectral graph anomaly detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Philadelphia: AAAI, 2025.
[39] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Advances in Neural Information Processing Systems. Long Beach: NeurIPS, 2017.
- 相似文献/References:
-
[1]丁永生.计算智能的新框架：生物网络结构[J].智能系统学报,2007,2(2):26.
DING Yong-sheng.A new scheme for computational intelligence: bio-network architecture[J].CAAI Transactions on Intelligent Systems,2007,2():26.
[2]徐 雄.人工情感的进化控制系统实现[J].智能系统学报,2008,3(2):135.
XU Xiong.Implementation of an evolutionary control system based on artificial emotion[J].CAAI Transactions on Intelligent Systems,2008,3():135.
[3]周孔丹,李 宁,鲁华祥.单电子电路的鲁棒性研究[J].智能系统学报,2008,3(3):195.
ZHOU Kong-dan,LI Ning,LU Hua-xiang.Researching the robustness of single electron devices[J].CAAI Transactions on Intelligent Systems,2008,3():195.
[4]张米娜,韩红桂,乔俊飞.前馈神经网络结构动态增长-修剪方法[J].智能系统学报,2011,6(2):101.
ZHANG Mina,HAN Honggui,QIAO Junfei.Research on dynamic feedforward neural network structure based on growing and pruning methods[J].CAAI Transactions on Intelligent Systems,2011,6():101.
[5]张文辉,高九州,马静,等.漂浮基空间机器人的径向基神经网络鲁棒自适应控制[J].智能系统学报,2011,6(2):114.
ZHANG Wenhui,GAO Jiuzhou,MA Jing,et al.The RBF neural network robust adaptive control of a freefloating space robot[J].CAAI Transactions on Intelligent Systems,2011,6():114.
[6]薄迎春,乔俊飞,杨刚.一种多模块协同参与的神经网络[J].智能系统学报,2011,6(3):225.
BO Yingchun,QIAO Junfei,YANG Gang.A multimodule cooperative neural network[J].CAAI Transactions on Intelligent Systems,2011,6():225.
[7]蒲兴成,张军,张毅.基于神经网络的改进行为协调控制及其在智能轮椅路径规划中的应用[J].智能系统学报,2011,6(5):456.
PU Xingcheng,ZHANG Jun,ZHANG Yi.Modified behavior coordination for intelligent wheelchair path planning based on a neural network[J].CAAI Transactions on Intelligent Systems,2011,6():456.
[8]段海庆,朱齐丹.基于反步自适应神经网络的船舶航迹控制[J].智能系统学报,2012,7(3):259.
DUAN Haiqing,ZHU Qidan.Trajectory tracking control of ships based onan adaptive backstepping neural network[J].CAAI Transactions on Intelligent Systems,2012,7():259.
[9]乔俊飞,逄泽芳,韩红桂.基于改进粒子群算法的污水处理过程神经网络优化控制[J].智能系统学报,2012,7(5):429.
QIAO Junfei,PANG Zefang,HAN Honggui.Neural network optimal control for wastewater treatment processbased on APSO[J].CAAI Transactions on Intelligent Systems,2012,7():429.
[10]郭一,刘金琨.带执行器饱和的柔性关节机器人位置反馈动态面控制[J].智能系统学报,2013,8(1):21.[doi:10.3969/j.issn.1673-4785.201204012]
GUO Yi,LIU Jinkun.Position feedback dynamic surface control for flexible joint robots with actuator saturation[J].CAAI Transactions on Intelligent Systems,2013,8():21.[doi:10.3969/j.issn.1673-4785.201204012]
备注/Memo
收稿日期:2025-10-27。
基金项目:国家重点研发计划项目（2019YFD0900900/05）；福建省教育厅中青年教师教育科研项目（JAT241070）；福建理工大学科技项目（GY-Z21183）.
作者简介:王天昊，硕士研究生，主要研究方向为人工智能和图神经网络。E-mail：2241308079@smail.fjut.edu.cn。;王翔，副教授，博士，主要研究方向为人工智能和图神经网络。CCF专业会员。 E-mail：wxsyhwl@fjut.edu.cn。;毛国君，教授，博士，主要研究方向为人工智能、数据挖掘、大数据和分布式计算。中国人工智能学会专委会常委、国家科学技术奖评审委员、计算机学会生信息物学专委会委员。发表学术论文100余篇。E-mail：19662092@fjut.edu.cn。
通讯作者:王翔. E-mail：wxsyhwl@fjut.edu.cn
更新日期/Last Update:
1900-01-01