Cross-Institution Cooperative Detection Model of Insurance Fraud Based on Graph Neural Networks and Federated Learning
Harvard Extension School, Harvard University, Cambridge, Massachusetts, United States
* Corresponding author: Chen Zhao. Email: mr.zhaoc@outlook.com
Abstract
The core challenge in cross-institution insurance fraud detection is that fraud-ring connectivity crosses institutional boundaries, while privacy regulations prohibit the centralized aggregation of raw claims data. This paper proposes CFGFraud, a federated graph detection framework for partially observable graphs. In this framework, a virtual-node placeholder mechanism makes the local model aware of cross-institutional hub entities; institutional attention mask separates the parameterization of real and virtual connections and imposes a learnable attenuation coefficient; and a federated attention alignment strategy constrains drift in each institution's attention pattern during aggregation. The non-IID experiment with five clients on the IEEE-CIS dataset shows that CFGFraud achieves an AUPRC of 0.791, 0.057 higher than FedAvg+GAT. Fraud-ring recall reaches 0.723, 24.4% higher than FedAvg+GAT. The standard deviation across clients is compressed from 0.029 to 0.012. Single-round computation costs increase by about 8%, while communication overhead increases by less than 0.01%. Ablation experiments show that the virtual-node mechanism accounts for 31.1% of the fraud-ring recall gain, and collaboration gain is positively correlated with entity overlap across institutions. This framework provides a deployable technical path for cross-agency relational fraud detection under the constraint of data confinement.
Keywords
Insurance fraud detection
Graph neural network
Federated learning
Cross-institution collaboration
Attention alignment
References
- Kumar, A. K., Nivethitha, T., Poonguzhali, P. K., & Saranyanandhini, D. (2026). Zero‐Knowledge Proofs for Model Integrity and Privacy. Decentralized Collaborative Learning for Data Privacy and Security, 211-249. DOI: 10.1002/9781394464746.ch10
- Mudannayake, O., Indika, A., Jayasinghe, U., Lee, G. M., & Alawatugoda, J. (2025). On privacy-preserved machine learning using secure multi-party computing: Techniques and trends. Computers, Materials & Continua, 85(2), 2527-2578. DOI: 10.32604/cmc.2025.068875
- Baduwal, M., Paudel, P., & Chaudhary, V. (2026). Federated learning: A survey of core challenges, current methods, and opportunities. Computers, 15(3), 155. DOI: 10.3390/computers15030155
- Yazdinejad, A., & Kong, J. D. (2026). Breaking interprovincial data silos: how federated learning can unlock Canada's public health potential. Royal Society Open Science, 13(3). DOI: 10.1098/rsos.250959
- Özdemir, N., & Gökçen, H. (2026, July). Interpretable Academic Team Formation on Heterogeneous Information Networks: Constructive Heuristics with Explicit Organizational Affiliation. In Informatics (Vol. 13, No. 7, p. 107). MDPI. DOI: 10.3390/informatics13070107
- Gross, C., & Siklos, P. L. (2020). Analyzing credit risk transmission to the nonfinancial sector in Europe: A network approach. Journal of Applied Econometrics, 35(1), 61-81. DOI: 10.1002/jae.2726
- Shan, C., Wang, C., & Wang, J. (2025, May). Regularized Incremental Federated Learning for Fraud Detection in Vehicle Insurances. In 2025 28th International Conference on Computer Supported Cooperative Work in Design (CSCWD) (pp. 1832-1837). IEEE. DOI: 10.1109/cscwd64889.2025.11033178
- Zhao, T., Chen, G., Pang, C., Seenoi, P., Papukdee, N., Busababodhin, P., & Du, Y. (2025). Hybrid convolutional neural network-graph attention network-gradient boosting decision tree model for seismic impedance inversion prediction. Journal of Seismic Exploration, 34(5), 81. DOI: 10.36922/JSE025310051
- Chen, G., Zhao, T., Pang, C., Seenoi, P., Papukdee, N., & Busababodhin, P. (2025). An attention-guided graph neural network and U-Net++-based reservoir porosity prediction system. Journal of Seismic Exploration, 34(4), 70. DOI: 10.36922/JSE025300044
- Jing, S., Zhang, L., & Wang, X. (2026). MM-TFDX: a multimodal graph neural network framework for dynamic cross-border tax fraud detection with regulation-aware explainability. Scientific Reports. DOI: 10.1038/s41598-026-64839-4
- Zhao, T., Chen, G., Li, J., Pang, C., Seenoi, P., Papukdee, N., & Busababodhin, P. (2026). XGBoost-Deep Residual (FedXGB-ResNet) collaborative porosity prediction in a federated learning framework. Journal of Seismic Exploration, 35(2), 201. DOI: 10.36922/JSE025440094
- Zeng, X. (2026). Cross-Border Trade Fraud Detection via Integrated Heterogeneous Graph Neural Network and XGBoost. International Scientific Technical and Economic Research, 4(1), 47-69. DOI: 10.71451/ISTAER2603
- Khan, W., Ishrat, M., Faisal, S. M., Ahmad, F., Nabilal, K. V., & Sharma, Y. K. (2025). A robust framework for topology-based anomaly detection in attributed networks using graph attention networks, substructure analysis, and data augmentation. International Journal of Data Science and Analytics, 20(7), 6715-6728. DOI: 10.1007/s41060-025-00848-2
- Wang, Y., Lei, J., Li, Y., & Shang, F. (2025). GATR: a graph attention deep reinforcement learning approach with variance-sensitive rewards for rpl routing optimization. Journal of King Saud University Computer and Information Sciences, 37(9), 274. DOI: 10.1007/s44443-025-00266-1
- Zhao, T., Chen, G., Pang, C., Seenoi, P., Papukdee, N., & Busababodhin, P. (2025). Time-lapse earthquake difference prediction based on physics-informed long short-term memory coupled with interpretability boosting. Journal of Seismic Exploration, 34(3), 25. DOI: 10.36922/JSE025310049
- Li, Y., & Ma, X. (2026). Intelligent delineation algorithm of urban development boundary based on graph neural network. International Scientific Technical and Economic Research, 4(1), 123-148. DOI: 10.71451/ISTAER2606
- Lin, X., Zhao, X., Yin, Z., Pan, H., Jing, P., Li, W., ... & Shu, Y. (2025). Knowledge-driven federated learning: A systematic literature review on approaches, challenges, and prospects: X. Lin et al. The Journal of Supercomputing, 81(8), 1016. DOI: 10.1007/s11227-025-07516-z
- Shen, G., Guo, X., Yuan, H., Du, J., Zhang, W., Qu, Y. N., & Kong, X. (2026). Communication-efficient decentralized federated graph learning via knowledge distillation under dual heterogeneity. Information Sciences, 123192. DOI: 10.1016/j.ins.2026.123192
- Yue, W., Lai, T., Mao, Q., Li, Q., & Camacho, D. (2026). A Review of Federated Learning Under Data Heterogeneity. Expert Systems, 43(6), e70271. DOI: 10.1111/exsy.70271
- Yan, C., Lu, X., Lio, P., Hui, P., & He, D. (2024). Self-Simulation and Meta-Model Aggregation-Based Heterogeneous-Graph-Coupled Federated Learning. IEEE Internet of Things Journal, 12(1), 198-212. DOI: 10.1109/JIOT.2024.3462724
- Yang, H., Wang, M., Dai, L., Wu, Y., & Du, J. (2025, August). Federated graph neural networks for heterogeneous graphs with data privacy and structural consistency. In 2025 5th International Conference on Computer Science and Blockchain (CCSB) (pp. 311-315). IEEE. DOI: 10.1109/ccsb66722.2025.11154261
- Hu, Y., & Yan, T. (2026). Beyond Simple Aggregation: Decoupling Frequency-Aware Representation Learning and Gradient Boosting for Graph Anomaly Detection. IEEE Access. DOI: 10.1109/access.2026.3714218
- Wang, X., Wiegand, T., Kraft, N. J., Swenson, N. G., Davies, S. J., Hao, Z., ... & Wolf, A. (2016). Stochastic dilution effects weaken deterministic effects of niche‐based processes in species rich forests. Ecology, 97(2), 347-360. DOI: 10.1890/14-2357.1
- Mukhoti, J., Kulharia, V., Sanyal, A., Golodetz, S., Torr, P., & Dokania, P. (2020). Calibrating deep neural networks using focal loss. Advances in Neural Information Processing Systems, 33, 15288-15299. DOI: 10.48550/arXiv.2002.09437
- Ma, J., Xiang, S., Li, Q., Yuan, L., Cheng, D., & Jiang, C. (2024). Parallel graph learning with temporal stamp encoding for fraudulent transactions detections. IEEE Transactions on Big Data, 11(4), 1945-1958. DOI: 10.1109/TBDATA.2024.3499338
- Devi, R. R., Raja, J. E., & Chin, Y. B. (2025). Reinforcement learning with graph neural network (RL-GNN) fusion for real-time financial fraud detection: a context-aware community mining approach. Scientific Reports, 15(1), 42953. DOI: 10.1038/s41598-025-25200-3
- Guo, J., Li, Z., Feng, X., Zhong, Z., & Bao, R. (2026). Label-Safe Temporal Graph Attention for Financial Fraud Detection in Transaction Networks. IEEE Access. DOI: 10.1109/access.2026.3752923