DPxFin: Adaptive Differential Privacy for Anti-Money Laundering Detection via Reputation-Weighted Federated Learning
arXiv:2603.19314v1 Announce Type: new Abstract: In the modern financial system, combating money laundering is a critical challenge complicated by data privacy concerns and increasingly complex fraud transaction patterns. Although federated learning (FL) is a promising problem-solving approach as it allows institutions to train their models without sharing their data, it has the drawback of being prone to privacy leakage, specifically in tabular data forms like financial data. To address this, we propose DPxFin, a novel federated framework that integrates reputation-guided adaptive differential privacy. Our approach computes client reputation by evaluating the alignment between locally trained models and the global model. Based on this reputation, we dynamically assign differential privacy noise to client updates, enhancing privacy while maintaining overall model utility. Clients with higher reputations receive lower noise to amplify their trustworthy contributions, while low-reputatio
arXiv:2603.19314v1 Announce Type: new Abstract: In the modern financial system, combating money laundering is a critical challenge complicated by data privacy concerns and increasingly complex fraud transaction patterns. Although federated learning (FL) is a promising problem-solving approach as it allows institutions to train their models without sharing their data, it has the drawback of being prone to privacy leakage, specifically in tabular data forms like financial data. To address this, we propose DPxFin, a novel federated framework that integrates reputation-guided adaptive differential privacy. Our approach computes client reputation by evaluating the alignment between locally trained models and the global model. Based on this reputation, we dynamically assign differential privacy noise to client updates, enhancing privacy while maintaining overall model utility. Clients with higher reputations receive lower noise to amplify their trustworthy contributions, while low-reputation clients are allocated stronger noise to mitigate risk. We validate DPxFin on the Anti-Money Laundering (AML) dataset under both IID and non-IID settings using Multi Layer Perceptron (MLP). Experimental analysis established that our approach has a more desirable trade-off between accuracy and privacy than those of traditional FL and fixed-noise Differential Privacy (DP) baselines, where performance improvements were consistent, even though on a modest scale. Moreover, DPxFin does withstand tabular data leakage attacks, proving its effectiveness under real-world financial conditions.
Executive Summary
This study proposes DPxFin, a novel federated learning framework that integrates reputation-guided adaptive differential privacy to address the challenge of money laundering detection in the financial system. The approach computes client reputation by evaluating the alignment between locally trained models and the global model, dynamically assigning differential privacy noise to client updates. Experimental analysis on the Anti-Money Laundering dataset demonstrated that DPxFin achieves a desirable trade-off between accuracy and privacy, outperforming traditional federated learning and fixed-noise differential privacy baselines. The framework also withstands tabular data leakage attacks, proving its effectiveness under real-world financial conditions. As a promising solution to the complex challenge of money laundering detection, DPxFin has significant implications for the financial industry and policymakers.
Key Points
- ▸ DPxFin integrates reputation-guided adaptive differential privacy to enhance privacy preservation in federated learning
- ▸ The framework computes client reputation by evaluating the alignment between locally trained models and the global model
- ▸ DPxFin achieves a desirable trade-off between accuracy and privacy, outperforming traditional federated learning and fixed-noise differential privacy baselines
Merits
Enhanced Privacy Preservation
DPxFin's adaptive differential privacy approach dynamically assigns noise to client updates based on client reputation, ensuring a more robust privacy preservation mechanism.
Improved Accuracy
By assigning lower noise to high-reputation clients, DPxFin amplifies trustworthy contributions and maintains overall model utility, leading to improved accuracy.
Effectiveness in Tabular Data
DPxFin withstands tabular data leakage attacks, demonstrating its effectiveness under real-world financial conditions.
Demerits
Scalability Limitations
The framework's performance improvements were modest, suggesting scalability limitations in handling large datasets or complex models.
Data Quality Requirements
DPxFin's reputation computation relies on the quality of locally trained models, which may not always be available or reliable.
Expert Commentary
DPxFin's innovative approach to reputation-guided adaptive differential privacy offers a promising solution to the complex challenge of money laundering detection. While the framework's performance improvements were modest, the study's findings demonstrate the potential of DPxFin in real-world financial applications. As the financial industry continues to grapple with the challenges of privacy preservation and model utility, DPxFin's adaptive differential privacy mechanism provides a valuable contribution to the field. Future research should focus on scaling DPxFin to handle large datasets and complex models, as well as exploring its application in other financial domains.
Recommendations
- ✓ Further research should focus on scaling DPxFin to handle large datasets and complex models
- ✓ Application of DPxFin in other financial domains, such as credit risk assessment and market analysis, should be explored
Sources
Original: arXiv - cs.LG
