Federated Learning-Based Framework for Privacy-Protected Cross-Border Financial Risk Evaluation: Analyzing US-Asia Investment Flows

Abstract

This paper presents a novel federated learning framework for privacy-preserving cross-border financial risk assessment, specifically focused on US-Asia investment flows. Cross-border financial transactions face significant challenges in risk assessment due to disparate regulatory environments, data sovereignty requirements, and privacy constraints across jurisdictions. Our proposed architecture addresses these challenges through a multi-layered approach that incorporates differential privacy, homomorphic encryption, and secure aggregation techniques while enabling collaborative model training without raw data exchange. Experimental results demonstrate that the proposed framework achieves 94.5% detection accuracy with 217ms latency in real-world case studies, outperforming conventional federated learning approaches by 4.3-7.2% across key performance metrics while maintaining regulatory compliance. The architecture reduces false positives by 73% compared to baseline methods while preserving data locality requirements. Privacy protection analysis confirms resilience against multiple attack vectors with only 0.4% model inversion success rate compared to 7.2% for state-of-the-art alternatives. This research establishes a foundation for enhanced cross-jurisdictional financial risk assessment that balances analytical capabilities with strict privacy preservation, enabling financial institutions to develop more sophisticated risk models across US and Asian markets without compromising regulatory compliance or data sovereignty.