基于局部差分与全局谱融合的图异常检测

doi:10.16451/j.cnki.issn1003-6059.202507005

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (3183 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要图神经网络在图异常检测任务中性能较优,但现有研究大多存在难以识别异常伪装、标签稀缺条件下性能不足等问题.针对上述问题,文中提出基于局部差分与全局谱融合的图异常检测方法(Graph Anomaly Detection Based on Local Differential and Global Spectral Fusion, GAD-LDSF),同时从空间域与频域进行特征嵌入,提升在异常伪装与标签稀缺条件下的检测性能.首先,利用基于软量化的概率自适应嵌入实现数据增强,通过图差分注意力网络精准捕获节点间的细微差异,生成空间域节点表示.然后,构建全局同构化动态频谱增强,利用Chebyshev多项式高效捕捉全局范围的异常信号,生成频域节点表示.最后,动态融合空间域与频域节点表示,并利用预测器实现异常节点检测.在3个基准数据集上的实验表明,GAD-LDSF整体性能较优,尤其是在应对异常伪装与标签稀缺等挑战时表现出较强的鲁棒性与泛化性.

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	徐登彬
	袁立宁
	吴沛宸
	刘钊

关键词 ：图异常检测, 图表示学习, 图神经网络, 注意力机制, 软量化

Abstract：Graph neural networks(GNNs) are widely recognized for their effectiveness in graph anomaly detection tasks. However, existing methods commonly struggle to identify camouflaged anomalies and exhibit poor performance under label scarcity. To address these issues, a method for graph anomaly detection based on local differential and global spectral fusion(GAD-LDSF) is proposed. Feature embeddings are employed in both spatial and spectral domains to enhance performance under camouflaged anomalies and label scarcity. First, probability-adaptive continuum embedding based on soft quantization is utilized to achieve data augmentation. A graph differential attention network is employed to accurately capture subtle differences between nodes and generate node representations in the spatial domain. Then, a globally homogenized dynamic spectral enhancement module is constructed. Chebyshev polynomials are leveraged to efficiently capture global anomaly signals and generate node representations in the spectral domain. Finally, spatial domain and spectral domain node representations are dynamically fused, and a predictor is employed to achieve anomalous node detection. Experiments on three benchmark datasets validate that GAD-LDSF achieves overall superior performance, particularly demonstrating its strong robustness and generalization in handling challenges such as camouflaged anomalies and label scarcity.

Key words： Graph Anomaly Detection Graph Representation Learning Graph Neural Networks Attention Mechanism Soft Quantization

收稿日期: 2025-04-29

ZTFLH:

TP391

基金资助:中国人民公安大学研究生科研创新项目(No.2025YJSKY012)、国家重点研发计划项目(No.2023YFC3321604)资助

通讯作者: 刘钊,博士,副研究员,主要研究方向为机器学习、计算机视觉.E-mail:liuzhao@ppsuc.edu.cn.

作者简介: 徐登彬,硕士研究生,主要研究方向为机器学习、图异常检测.E-mail:2023211471@stu.ppsuc.edu.cn.袁立宁,博士研究生,高级工程师,主要研究方向为机器学习、图神经网络.E-mail:yuanlining@gxjcxy.edu.cn.吴沛宸,硕士,主要研究方向为机器学习、计算机视觉.E-mail:m13209406252@163.com.

引用本文:

徐登彬, 袁立宁, 吴沛宸, 刘钊. 基于局部差分与全局谱融合的图异常检测[J]. 模式识别与人工智能, 2025, 38(7): 641-654. XU Dengbin, YUAN Lining, WU Peichen, LIU Zhao. Graph Anomaly Detection Based on Local Differential and Global Spectral Fusion. Pattern Recognition and Artificial Intelligence, 2025, 38(7): 641-654.

链接本文:

http://manu46.magtech.com.cn/Jweb_prai/CN/10.16451/j.cnki.issn1003-6059.202507005 或 http://manu46.magtech.com.cn/Jweb_prai/CN/Y2025/V38/I7/641

[1] 胡威,李姝影,张鑫,等.基于图与语义表示学习的专利引文网络链路预测研究.数据分析与知识发现, 2024, 8(10): 28-43.
(HU W, LI S Y, ZHANG X, et al. Link Prediction in Patent Citation Networks Based on Graph and Semantic Representation Lear-ning. Data Analysis and Knowledge Discovery, 2024, 8(10): 28-43.)
[2] 代星月,叶海良,曹飞龙.社交影响增强的图神经网络推荐方法.模式识别与人工智能, 2024, 37(3): 221-230.
(DAI X Y, YE H L, CAO F L. Graph Neural Network Recommendation Based on Enhanced Social Influence. Pattern Recognition and Artificial Intelligence, 2024, 37(3): 221-230.)
[3] YING R, HE R N, CHEN K F, et al. Graph Convolutional Neural Networks for Web-Scale Recommender Systems // Proc of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2018: 974-983.
[4] XIANG S, ZHU M Z, CHENG D W, et al. Semi-supervised Credit Card Fraud Detection via Attribute-Driven Graph Representation. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(12): 14557-14565.
[5] 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.计算机研究与发展, 2021, 58(7): 1436-1455.
(CHEN B F, LI J D, LU X J, et al. Survey of Deep Learning Based Graph Anomaly Detection Methods. Journal of Computer Research and Development, 2021, 58(7):1436-1455)
[6] GORI M, MONFARDINI G, SCARSELLI F. A New Model for Lear-ning in Graph Domains // Proc of the IEEE International Joint Conference on Neural Networks. Washington, USA: IEEE, 2005: 729-734.
[7] ZHANG S, TONG H H, XU J J, et al. Graph Convolutional Networks: A Comprehensive Review. Computational Social Networks, 2019, 6. DOI: 10.1186/s40649-019-0069-y.
[8] LIU Y, AO X, QIN Z D, et al. Pick and Choose: A GNN-Based Imbalanced Learning Approach for Fraud Detection // Proc of the Web Conference. New York, USA: ACM, 2021: 3168-3177.
[9] CHANG W J, YU J J, ZHOU X J. ConsE: Consistency Exploitation for Semi-supervised Anomaly Detection in Graphs // Proc of the International Joint Conference on Neural Networks. Washington, USA: IEEE, 2023. DOI: 10.1109/IJCNN54540.2023.10191356.
[10] HYUN W, LEE I, SUH B. LEX-GNN: Label-Exploring Graph Neu-ral Network for Accurate Fraud Detection // Proc of the 33rd ACM International Conference on Information and Knowledge Management. New York, USA: ACM, 2024: 3802-3806.
[11] DOU Y T, LIU Z W, SUN L, et al. Enhancing Graph Neural Network-Based Fraud Detectors Against Camouflaged Fraudsters // Proc of the 29th ACM International Conference on Information and Knowledge Management. New York, USA: ACM, 2020: 315-324.
[12] LIU Z W, DOU Y T, YU P S, et al. Alleviating the Inconsistency Problem of Applying Graph Neural Network to Fraud Detection // Proc of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. New York, USA: ACM, 2020: 1569-1572.
[13] KIM H, CHOI J, WHANG J J. Dynamic Relation-Attentive Graph Neural Networks for Fraud Detection // Proc of the IEEE International Conference on Data Mining Workshops. Washington, USA: IEEE, 2023: 1092-1096.
[14] ZHAO J, SHAO M L, TANG H L, et al. RHGNN: Fake Reviewer Detection Based on Reinforced Heterogeneous Graph Neural Networks. Knowledge-Based Systems, 2023, 280. DOI: 10.1016/j.knosys.2023.111029.
[15] TANG J H, LI J J, GAO Z Q, et al. Rethinking Graph Neural Networks for Anomaly Detection. Proceedings of Machine Learning Research, 2022, 162: 21076-21089.
[16] GAO Y, WANG X, HE X N, et al. Addressing Heterophily in Graph Anomaly Detection: A Perspective of Graph Spectrum // Proc of the ACM Web Conference. New York, USA: ACM, 2023: 1528-1538.
[17] YE T Z, DONG L, XIA Y Q, et al. Differential Transformer[C/OL].[2025-03-21]. https://arxiv.org/pdf/2410.05258.
[18] VELIČKOVIČ P, CUCURULL G, CASANOVA A, et al. Graph Attention Networks[C/OL].[2025-03-21]. https://arxiv.org/pdf/1710.10903.
[19] DEFFERRARD M, BRESSON X, VANDERGHEYNST P. Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering // Proc of the 30th International Conference on Neural Information Processing Systems. Cambridge, USA: MIT Press, 2016: 3844-3852.
[20] RAYANA S, AKOGLU L. Collective Opinion Spam Detection: Bri-dging Review Networks and Metadata // Proc of the 21st ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2015: 985-994.
[21] WU B, YAO X Y, ZHANG B Y, et al. SplitGNN: Spectral Graph Neural Network for Fraud Detection Against Heterophily // Proc of the 32nd ACM International Conference on Information and Know-ledge Management. New York, USA: ACM, 2023: 2737-2746.
[22] LI Y Q, ZANG G S, SONG C Y, et al. A Universal Adaptive Algorithm for Graph Anomaly Detection. Information Processing and Management, 2025, 62(1). DOI: 10.1016/j.ipm.2024.103905.
[23] WANG X, DOU H, DONG D B, et al. Graph Anomaly Detection Based on Hybrid Node Representation Learning. Neural Networks, 2025, 185. DOI: 10.1016/j.neunet.2025.107169.