Sep 2026· EAI Endorsed Transactions on Scalable Information Systems· Vol 13· 0 citations· 35 references
Computer Science
TL;DR
The results indicate that the proposed edge-driven federated learning framework can support privacy-preserving and robust cross-institutional financial risk modeling and provides an effective solution for collaborative fraud detection and anti-money laundering under data isolation, heterogeneous edge environments, and adversarial conditions.
Abstract
INTRODUCTION: Cross-institutional collaborative financial risk modeling can improve fraud detection and anti-money laundering performance. However, data sensitivity, institutional isolation, non-IID data distributions, and heterogeneous edge nodes make traditional federated learning difficult in terms of privacy protection, communication efficiency, and robustness.
Objectives
This paper aims to develop an edge-driven federated learning framework for secure cross-institutional financial risk modeling, enabling collaborative learning without sharing raw financial data.
Methods
The proposed framework integrates risk-relationship-aware dynamic grouping, privacy-constrained representation transmission, robust secure aggregation, and low-dimensional summary consistency checking. By limiting the scope of information exchange and controlling representation sharing, the framework reduces representation deviation and mitigates the influence of abnormal updates.
Results
Experiments were conducted on five public or synthetic financial risk datasets. The proposed method achieved the best federated AUPRC on four datasets and was only slightly lower than DSFL on the Elliptic dataset. Under client dropout and malicious-node settings, the method maintained relatively stable AUPRC performance and showed favorable communication efficiency and robustness.
Conclusion
The results indicate that the proposed edge-driven federated learning framework can support privacy-preserving and robust cross-institutional financial risk modeling. It provides an effective solution for collaborative fraud detection and anti-money laundering under data isolation, heterogeneous edge environments, and adversarial conditions.
A governance-constrained federated learning model for cross-regional data sharing that enables the global model to adjust aggregation weights, privacy intensity, and transmission frequency according to node heterogeneity, data sensitivity, and governance authorization status is developed.
The proposed model effectively improves the efficiency and timeliness of collaborative detection of cross-organizational threats while ensuring data privacy and provides a feasible solution for building a safe and reliable collaborative defense system.
SecureFedShield is proposed, a privacy-preserving federated learning framework designed for secure financial fraud detection in adversarial environments that integrates adaptive privacy protection, trust-aware client evaluation, adversarial update detection, and robust model aggregation into a unified architecture.
Kriti Mishra· International Journal of Cre...· 0 citations
FedHDL (Federated Heterogeneous Deep Learning), a novel privacy-preserving framework for cryptocurrency fraud detection that enables collaborative model training across five heterogeneous institutional nodes without raw data exchange, is introduced.
Kanika Singhal· Journal of Intelligent Decis...· 0 citations
FedE, a multi-precision, multi-source, heterogeneous privacy-preserving federated learning training method based on functional encryption that enhances numerical adaptation during ciphertext computation and prevents model parameter updates from easily compromising privacy in cross-institutional federated learning.
Weijia Liu, Junwen Deng, Hao Li et al.· Computers, Materials & C...· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 17, 2026
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.
MIT News · Artificial Intelligence· news.mit.eduNov 17, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 9, 2026