One-Class Learning for Graph Neural Networks

• Marcos Gôlo USP
• Ricardo Marcacini USP

DOI: https://doi.org/10.5753/webmedia_estendido.2026.35006

Resumo

One-Class Learning (OCL) addresses scenarios in which only examples from a class of interest are available for training, while the remaining data space is broad, unknown, or continuously evolving. This thesis advances OCL for graphs through the OLGA family, four end-to-end Graph Neural Network (GNN) methods that progressively address node classification, graph streams, causal edge classification, and heterogeneous graphs. OLGA introduces the methodological core by combining a novel sphere-based loss with graph reconstruction in an interpretable (2D/3D) latent space. OLGASTREAM extends this formulation to evolving graphs, ECHOLGA adapts it to causal relations, and FOLGA exploits heterogeneous information through topology rewiring and fusion. Across 19 datasets from textual, image, tabular, streaming, causal, and heterogeneous domains, the proposed methods achieved the best 𝑓1-macro in 14 datasets. The results demonstrate the applicability of one-class GNNs to open-domain Web and multimedia problems in which the target concept is known, but the negative space cannot be exhaustively modeled.

Palavras-chave: One-Class Classification, Networks, Sphere Loss

Referências

Aoge Chen, Jianshe Wu, and Hongtao Zhang. 2025. FIAD: Graph anomaly detection framework based feature injection. Expert Systems with Applications (2025).

Fatemeh Deldar, Mahdi Abadi, and Mohammad Ebrahimifard. 2022. Android Malware Detection Using One-Class Graph Neural Networks. ISeCure (2022).

Frank Emmert-Streib and Matthias Dehmer. 2022. Taxonomy of machine learning paradigms: A data-centric perspective. Wiley Interdisciplinary Reviews: DM and KD (2022).

Haoyi Fan, Ruidong Wang, Xunhua Huang, Fengbin Zhang, Zuoyong Li, and Shimei Su. 2024. Deep joint adversarial learning for anomaly detection on attribute networks. Inf. Sci. (2024).

Yong Feng, Jinglong Chen, Zijun Liu, Haixin Lv, and Jun Wang. 2022. Full Graph Autoencoder for One-Class Group Anomaly Detection of IIoT System. Internet of Things Journal (2022).

Tom Ganz, Inaam Ashraf, Martin Härterich, and Konrad Rieck. 2023. Detecting Backdoors in Collaboration Graphs of Software Repositories. In Data and Application Security and Privacy.

Jingtao Hu, Bin Xiao, Hu Jin, Jingcan Duan, Siwei Wang, Zhao Lv, Siqi Wang, Xinwang Liu, and En Zhu. 2023. SAMCL: Subgraph-Aligned Multiview Contrastive Learning for Graph Anomaly Detection. Transactions on NN and Learning Systems (2023).

Bin Huang, Xuhong Wang, Ping Cui, Wenjian Jiang, Yupu Yang, and Qifu Fan. 2021. One-class Temporal Graph Attention Neural Network for Dynamic Graph Anomaly Detection. In ECIT.

Zeqi Huang, Yonghao Gu, and Qing Zhao. 2022. One-Class Directed Heterogeneous Graph Neural Network for Intrusion Detection. In Proc. Conf. Innovation in Artificial Intelligence.

Bharti Khemani, Shruti Patil, Ketan Kotecha, and Sudeep Tanwar. 2024. A review of graph neural networks: concepts, architectures, techniques, challenges, datasets, applications, and future directions. Journal of Big Data (2024).

Qing Li, Guanzhong Wu, Hang Ni, and Tao You. 2024. Anomaly detection with dual-channel heterogeneous graph based on hypersphere learning. Information Sciences (2024).

Bin Shi, Bo Dong, Yiming Xu, Jiaxiang Wang, Yunfan Wang, and Qinghua Zheng. 2023. An edge feature aware heterogeneous graph neural network model to support tax evasion detection. Expert Systems with Applications (2023).

Xuhong Wang, Baihong Jin, Ying Du, Ping Cui, Yingshui Tan, and Yupu Yang. 2021. One-class graph neural networks for anomaly detection in attributed networks. Neural C. and Applic. (2021).

Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and S Yu Philip. 2020. A comprehensive survey on graph neural networks. Trans.s on NN and Learning Systems (2020).

Fengbin Zhang, Haoyi Fan, Ruidong Wang, Zuoyong Li, and Tiancai Liang. 2022. Deep dual support vector data description for anomaly detection on attributed networks. International Journal of Intelligent Systems 37, 2 (2022), 1509–1528.

Junjie Zhang, Xinye Wang, Zhenyang Yu, Menglin Huang, Lei Duan, and Jie Zuo. 2022. STEAM: SVDD-based Anomaly Detection on Attributed Networks via Multi-Autoencoders. In Proc. Conf. High Performance Computing and Communications.