隨著網路與科技的發展，越來越多的電商平台、社群媒體興起，逐漸成為人們 生活中不可或缺的一部份。也正因如此，許多惡意用戶趁著這樣的機會，在不同的 平台之間為了謀取利益而做出傷害一般使用者權益的行為。為了保障用戶的安全， 人們開始投入異常用戶偵測的研究領域，以盡早偵測出惡意用戶並阻止他們的行為。
　　在本次的研究中，我們的目標是在動態雙分圖(Dynamic Bipartite Graph)上進行 非監督式異常用戶的偵測，並且提出嶄新的框架，名為三元變分異常檢測(TTGAE)。 我們透過新穎的前處理技術將動態雙分圖建構為使用者-互動-項目三分圖，並結合變 分圖自編碼器(Variational Graph Auto-Encoders) 以及不同的訓練目標和策略，使得模 型能夠學習到原本動態雙分圖中隱含的資訊，進而偵測出異常的使用者。
　　我們進行了多種不同的實驗，確認TTGAE在面對不同場景的動態雙分圖時，都 能夠以良好的性能偵測出潛在的異常用戶。並透過後續的消融實驗驗證本次提出來 的各個模組都能夠有效的提升TTGAE的性能，確認了所提出的方法之有效性。

The proliferation of e-commerce platforms and social media has created digital ecosystems that are increasingly vulnerable to exploitation by malicious actors seeking personal gain at the expense of legitimate users. This growing security concern necessitates effective anomalous user detection mechanisms capable of early identification and mitigation of harmful activities across diverse online platforms. This research addresses the challenge of unsupervised anomalous user detection in dynamic bipartite graphs through a novel framework called Tripartite Temporal Graph Autoencoder (TTGAE). Our approach transforms conventional dynamic bipartite structures into user-interaction-item tripartite representations through innovative preprocessing techniques. By integrating Variational Graph Auto-Encoders with specialized reconstruction objectives and differentiated training strategies, TTGAE effectively extracts latent patterns from temporal interaction sequences for anomaly identification. Comprehensive experimental evaluation demonstrates that TTGAE achieves superior detection performance across multiple dynamic bipartite graph contexts. Through systematic ablation studies, we establish the significant contribution of each architectural component to the framework's overall effectiveness, validating our methodological innovations in unsupervised anomaly detection.

[1] Abien Fred Agarap. Deep learning using rectified linear units (relu), 2019.
[2] Sambaran Bandyopadhyay, Lokesh N, Saley Vishal Vivek, and M. N. Murty. Outlier resistant unsupervised deep architectures for attributed network embedding. InProceedings of the 13th International Conference on Web Search and Data Mining, WSDM’20, page 25–33, New York, NY, USA, 2020. Association for Computing Machinery.
[3] Markus M. Breunig, Hans-Peter Kriegel, Raymond T. Ng, and Jörg Sander. Lof: identifying density-based local outliers. SIGMOD Rec., 29(2):93–104, May 2000.
[4] Xiangyu Dong, Xingyi Zhang, Lei Chen, Mingxuan Yuan, and Sibo Wang. Spacegnn: Multi-space graph neural network for node anomaly detection with extremely limited labels, 2025.
[5] Yingtong Dou, Zhiwei Liu, Li Sun, Yutong Deng, Hao Peng, and Philip S. Yu. Enhancing graph neural network-based fraud detectors against camouflaged fraudsters. In Proceedings of the 29th ACM International Conference on Information &;Knowledge Management, CIKM ＇20, page 315–324. ACM, October 2020.
[6] Yingtong Dou, Kai Shu, Congying Xia, Philip S. Yu, and Lichao Sun. User preference aware fake news detection, 2021.
[7] Lanting Fang, Kaiyu Feng, Jie Gui, Shanshan Feng, and Aiqun Hu. Anonymous edge representation for inductive anomaly detection in dynamic bipartite graph. Proc. VLDB Endow., 16(5):1154–1167, January 2023.
[8] Rizal Fathony, Jenn Ng, and Jia Chen. Interaction-focused anomaly detection on bipartite node-and-edge-attributed graphs. In 2023 International Joint Conference on Neural Networks (IJCNN), pages 1–10, 2023.
[9] Adam Goodge, Bryan Hooi, See Kiong Ng, and Wee Siong Ng. Lunar: Unifying local outlier detection methods via graph neural networks, 2021.
[10] Bryan Hooi, Neil Shah, Alex Beutel, Stephan Gunnemann, Leman Akoglu, Mohit Kumar, Disha Makhija, and Christos Faloutsos. Birdnest: Bayesian inference for ratings fraud detection, 2016.
[11] Diederik P Kingma and Max Welling. Auto-encoding variational bayes, 2022.
[12] Thomas N. Kipf and Max Welling. Variational graph auto-encoders, 2016.
[13] Thomas N. Kipf and Max Welling. Semi-supervised classification with graph convolutional networks, 2017.
[14] Srijan Kumar, Xikun Zhang, and Jure Leskovec. Predicting dynamic embedding trajectory in temporal interaction networks. In Proceedings of the 25th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2019.
[15] Jinghan Li, Yuan Gao, Jinda Lu, Junfeng Fang, Congcong Wen, Hui Lin, and Xiang Wang. Diffgad: A diffusion-based unsupervised graph anomaly detector, 2025.
[16] Zheng Li, Yue Zhao, Nicola Botta, Cezar Ionescu, and Xiyang Hu. Copod: Copula based outlier detection. In 2020 IEEE International Conference on Data Mining (ICDM), page 1118–1123. IEEE, November 2020.
[17] Zheng Li, Yue Zhao, Xiyang Hu, Nicola Botta, Cezar Ionescu, and George H. Chen.Ecod:Unsupervisedoutlierdetectionusingempiricalcumulativedistributionfunctions. IEEE Transactions on Knowledge and Data Engineering, 35(12):12181–12193, December 2023.
[18] Zongwei Li, Lianghao Xia, Hua Hua, Shijie Zhang, Shuangyang Wang, and Chao Huang. Diffgraph: Heterogeneous graph diffusion model. In Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining, WSDM ’25, pages 40–49, New York, NY, USA, 2025. Association for Computing Machinery.
[19] Fei TonyLiu, Kai MingTing, andZhi-HuaZhou. Isolation forest. In 2008 Eighth IEEE International Conference on Data Mining, pages 413–422, 2008.
[20] Jiaxun Liu, Yue Tian, and Guanjun Liu. Global confidence degree based graph neural network for financial fraud detection, 2024.
[21] Kay Liu, Yingtong Dou, Xueying Ding, Xiyang Hu, Ruitong Zhang, Hao Peng, Lichao Sun, and Philip S. Yu. PyGOD: A Python library for graph outlier detection. Journal of Machine Learning Research, 25(141):1–9, 2024.
[22] Yen-Wen Lu, Yu-Che Tsai, and Cheng-Te Li. Burstiness-aware bipartite graph neural networks for fraudulent user detection on rating platforms. In Companion Proceedings of the ACM Web Conference 2024, WWW ’24, page 834–837, New York, NY, USA, 2024. Association for Computing Machinery.
[23] Xiaoxiao Ma, Fanzhen Liu, Jia Wu, Jian Yang, Shan Xue, andQuanZ.Sheng. Rethinking unsupervised graph anomaly detection with deep learning: Residuals and objectives. IEEE Transactions on Knowledge and Data Engineering, 37(2):881–895, 2025.
[24] Xiaoxiao Ma, Jia Wu, Shan Xue, Jian Yang, Chuan Zhou, Quan Z. Sheng, Hui Xiong, and Leman Akoglu. A comprehensive survey on graph anomaly detection with deep learning. IEEE Transactions on Knowledge and Data Engineering, 35(12):1201212038, December 2023.
[25] Junjun Pan, Yixin Liu, Yizhen Zheng, and Shirui Pan. Prem: A simple yet effective approach for node-level graph anomaly detection, 2023.
[26] Zhen Peng, Yunfan Wang, Qika Lin, Bo Dong, and Chao Shen. When bipartite graph learning meets anomaly detection in attributed networks: Understand abnormalities from each attribute. Neural Networks, 185:107194, 2025.
[27] Bryan Perozzi and Leman Akoglu. Discovering communities and anomalies in attributed graphs: Interactive visual exploration and summarization. ACM Trans. Knowl. Discov. Data, 12(2), January 2018.
[28] Bryan Perozzi, Leman Akoglu, Patricia Iglesias Sánchez, and Emmanuel Müller. Focused clustering and outlier detection in large attributed graphs. In Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’14, page 1346–1355, New York, NY, USA, 2014. Association for Computing Machinery.
[29] Amit Roy, Juan Shu, Jia Li, Carl Yang, Olivier Elshocht, Jeroen Smeets, and Pan Li. Gad-nr: Graph anomaly detection via neighborhood reconstruction. In Proceedings of the 17th ACMInternational Conference on Web Search and Data Mining, WSDM ＇24, page 576–585. ACM, March 2024.
[30] Tarek Seghair, Olfa Besbes, Takoua Abdellatif, and Sami Bihiri. Vq-vgae: Vectorquantized variational graph auto-encoder for unsupervised anomaly detection. In 2024 IEEE International Conference on Big Data (BigData), pages 2370–2375, 2024.
[31] QiuwuSha, TengdaTang, Xinyu Du, Jie Liu, Yixian Wang, and Yuan Sheng. Detecting credit card fraud via heterogeneous graph neural networks with graph attention. In 2025 IEEE 6th International Seminar on Artificial Intelligence, Networking and Information Technology (AINIT), pages 1332–1336, 2025.
[32] Moirangthem Tiken Singh, Rabinder Kumar Prasad, Gurumayum Robert Michael, N K Kaphungkui, and N. Hemarjit Singh. Heterogeneous graph auto-encoder for creditcard fraud detection, 2024.
[33] Shantanu Thakoor, Corentin Tallec, Mohammad Gheshlaghi Azar, Mehdi Azabou, Eva L. Dyer, Rémi Munos, Petar Veličković, and Michal Valko. Large-scale representation learning on graphs via bootstrapping, 2023.
[34] Andrew Wang, Rex Ying, Pan Li, Nikhil Rao, Karthik Subbian, and Jure Leskovec. Bipartite dynamic representations for abuse detection. In Proceedings of the 27th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2021.
[35] Xinzhi Wang, Jiayu Guo, Xiangfeng Luo, and Hang Yu. Dyhdge: Dynamic heterogeneous transaction graph embedding for safety-centric fraud detection in financial scenarios. Journal of Safety Science and Resilience, 5(4):486–497, 2024.
[36] Xin Xiong, Xiangyu Wang, Suorong Yang, Furao Shen, and Jian Zhao. Gmni: Achieve good data augmentation in unsupervised graph contrastive learning. Neural Netw., 181(C), January 2025.
[37] Bo Xu, Jinpeng Wang, Zhehuan Zhao, Hongfei Lin, and Feng Xia. Unsupervised anomaly detection on attributed networks with graph contrastive learning for consumer electronics security. IEEE Transactions on Consumer Electronics, 70(1):4062–4072, 2024.
[38] DaXu,Chuanwei Ruan, Evren Korpeoglu, Sushant Kumar, and Kannan Achan. Inductive representation learning on temporal graphs, 2020.
[39] Hongzuo Xu, Guansong Pang, Yijie Wang, and Yongjun Wang. Deep isolation forest for anomaly detection. IEEE Transactions on Knowledge and Data Engineering, 35(12):12591–12604, December 2023.
[40] Zhiming Xu, Xiao Huang, Yue Zhao, Yushun Dong, and Jundong Li. Contrastive attributed network anomaly detection with data augmentation. In Advances in Knowledge Discovery and Data Mining: 26th Pacific-Asia Conference, PAKDD 2022, Chengdu, China, May 16–19, 2022, Proceedings, Part II, page 444–457, Berlin, Heidelberg, 2022. Springer-Verlag.
[41] Yihang Yin, Qingzhong Wang, Siyu Huang, Haoyi Xiong, and Xiang Zhang. Autogcl: Automated graph contrastive learning via learnable view generators, 2022.
[42] Fengbin Zhang, Haoyi Fan, Ruidong Wang, Zuoyong Li, and Tiancai Liang. Deep dual support vector data description for anomaly detection on attributed networks, 2021.
[43] Ge Zhang, Jia Wu, Jian Yang, Amin Beheshti, Shan Xue, Chuan Zhou, and Quan Z. Sheng. Fraudre: Fraud detection dual-resistant to graph inconsistency and imbalance. In 2021 IEEE International Conference on DataMining (ICDM), pages 867–876, 2021.
[44] Yue Zhao, Xiyang Hu, Cheng Cheng, Cong Wang, Changlin Wan, Wen Wang, Jianing Yang, Haoping Bai, Zheng Li, Cao Xiao, Yunlong Wang, Zhi Qiao, Jimeng Sun, and Leman Akoglu. Suod: Accelerating large-scale unsupervised heterogeneous outlier detection, 2021.
[45] Yue Zhao, Zain Nasrullah, and Zheng Li. Pyod: A python toolbox for scalable outlier detection. Journal of Machine Learning Research, 20(96):1–7, 2019.
[46] Xiangping Zheng, Bo Wu, Xun Liang, and Wei Li. Friend or foe? mining suspicious behavior via graph capsule infomax detector against fraudsters. In Proceedings of the ACM Web Conference 2024, WWW’24, page 2684–2693, New York, NY, USA, 2024. Association for Computing Machinery.