https://doi.org/10.1140/epjds/s13688-025-00584-4
Research
A cross-social platform distributed anomaly behavior detection method
1
College of Information Engineering, Henan University of Science and Technology, Kaiyuan Avenue, 471000, Luoyang, Henan, China
2
College of Information Engineering, Southwest University of Science and Technology, Qinglong Avenue, 621010, Mianyang, Sichuan, China
a
This email address is being protected from spambots. You need JavaScript enabled to view it.
Received:
7
April
2025
Accepted:
21
August
2025
Published online:
29
August
2025
Abstract
With the rapid development of social networks, anomaly behavior detection has become essential for ensuring platform security and enhancing user experience. Traditional anomaly detection methods often rely on centralized data processing, making it difficult to address the challenges of cross-platform collaborative detection and failing to fully leverage the temporal and graph structure information within social networks. To improve anomaly behavior detection’s accuracy and generalization ability, this paper proposes a Cross-Platform-Based Distributed Anomaly Behavior Detection Method, FLAD. Specifically, the method employs the Federated Averaging (FedAvg) algorithm for model aggregation, constructing a decentralized anomaly behavior detection model. This approach avoids the exchange of raw user behavior data and enhances the effectiveness of cross-platform collaborative learning. Moreover, this paper introduces a detection model that combines a sliding window and Graph Convolutional Network (GCN), utilizing temporal data and the graph structure of social networks. By partitioning user behavior data into subgraphs via a sliding window, and employing GCN and Long Short-Term Memory (LSTM) models to learn the evolution patterns of temporal and behavioral data, the method improves the accuracy of anomaly behavior detection. Experimental results show that the proposed method achieves up to a 6.31% increase in F1-score compared to baseline models on the Epinions and Digg datasets. Ablation studies further demonstrate that the unified model, aggregated through federated learning, significantly improves accuracy over the initial global model. In the Epinions dataset, accuracy increased by 20.49%, and in the Digg dataset, accuracy improved by 23.13%.
Key words: Anomaly Behavior Detection / Cross-Platform Social Networks / Federated Learning / Graph Convolutional Network (GCN)
Handling Editor: Sabrina Gaito
Shiyu Li, Honghai Wu, Qi Zhang, Huahong Ma and Kaikai Deng contributed equally to this work.
© The Author(s) 2025
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
