Magnitude-Contrastive Network for Unsupervised Graph Anomaly Detection
Paper
Ge, Fei, Zhang, Ji, Wang, Zhen, Zhou, Yuqian and Li, Zhao. 2024. "Magnitude-Contrastive Network for Unsupervised Graph Anomaly Detection." H., Zhang W.Yang Z.Wang X.Tung A.Zheng Z.Guo (ed.) 8th APWeb-WAIM joint International Conference on Web and Big Data (APWeb-WAIM 2024). Jinhua, China 30 Aug - 01 Sep 2024 Singapore . Springer. https://doi.org/10.1007/978-981-97-7241-4_30
Paper/Presentation Title | Magnitude-Contrastive Network for Unsupervised Graph Anomaly Detection |
---|---|
Presentation Type | Paper |
Authors | Ge, Fei, Zhang, Ji, Wang, Zhen, Zhou, Yuqian and Li, Zhao |
Editors | H., Zhang W.Yang Z.Wang X.Tung A.Zheng Z.Guo |
Journal or Proceedings Title | Proceedings of the 8th APWeb-WAIM joint International Conference on Web and Big Data (APWeb-WAIM 2024) |
Journal Citation | 14964, pp. 481-493 |
Number of Pages | 13 |
Year | 2024 |
Publisher | Springer |
Place of Publication | Singapore |
ISBN | 9789819772407 |
9789819772414 | |
Digital Object Identifier (DOI) | https://doi.org/10.1007/978-981-97-7241-4_30 |
Web Address (URL) of Paper | https://link.springer.com/chapter/10.1007/978-981-97-7241-4_30 |
Web Address (URL) of Conference Proceedings | https://link.springer.com/book/10.1007/978-981-97-7241-4 |
Conference/Event | 8th APWeb-WAIM joint International Conference on Web and Big Data (APWeb-WAIM 2024) |
Event Details | 8th APWeb-WAIM joint International Conference on Web and Big Data (APWeb-WAIM 2024) Parent APWeb-WAIM International Joint Conference on Web and Big Data Delivery In person Event Date 30 Aug 2024 to end of 01 Sep 2024 Event Location Jinhua, China |
Abstract | Effectively identifying anomalous nodes within networks is crucial for various applications, such as fraud detection, network intrusion prevention, and social network activity monitoring. Existing graph anomaly detection methods based on contrastive learning have shown promising results but often suffer from limitations due to their local focus, such as community blindness (overlooking the inherent community structure of networks), limited anomaly scope (focusing solely on the local perspective), and high computational cost. To address these challenges, we propose a novel and efficient graph anomaly detection method called MCEN-GAD, which leverages a multi-level contrastive learning approach to identify anomalies across multiple network dimensions. Specifically, MCEN-GAD incorporates three contrastive networks: a patch-level contrastive network for local anomaly detection, a community-level contrastive network for identifying anomalies within specific communities, and a global-level anomaly detection network for exploring more global anomalous information. MCEN-GAD integrates the anomaly scores from these three levels using a weighted sum approach, achieving a comprehensive understanding of anomalous activity within the network. This multi-level integration allows MCEN-GAD to effectively capture anomalies across different network dimensions and provide a more robust anomaly detection framework. The experimental results clearly demonstrate the remarkable effectiveness and efficiency of our method compared with the state-of-the-art approaches on six benchmark datasets. |
Keywords | Anomaly Detection; Contrastive Learning; Unsupervised Learning |
Contains Sensitive Content | Does not contain sensitive content |
ANZSRC Field of Research 2020 | 460999. Information systems not elsewhere classified |
Public Notes | Files associated with this item cannot be displayed due to copyright restrictions. |
Series | Lecture Notes in Computer Science |
Byline Affiliations | Nanjing University of Aeronautics and Astronautics, China |
University of Southern Queensland | |
Zhejiang Lab, China | |
Hangzhou Yugu Technology, China |
Permalink -
https://research.usq.edu.au/item/z9958/magnitude-contrastive-network-for-unsupervised-graph-anomaly-detection
Download files
15
total views9
total downloads8
views this month3
downloads this month