preprints.org > business, economics and management > business and management > doi: 10.20944/preprints202409.1816.v1

Preprint Article Version 1 Preserved in Portico This version is not peer-reviewed

Version 1 : Received: 22 September 2024 / Approved: 23 September 2024 / Online: 24 September 2024 (05:09:10 CEST)

Federated unlearning represents a sophisticated evolution in the domain of machine learning, particularly within federated learning frameworks. In financial applications, where data privacy and security are paramount, federated unlearning allows institutions to selectively remove or “unlearn” specific data from trained models without compromising their overall performance and accuracy. This capability is essential for ensuring that models remain adaptable, secure, and compliant with regulatory requirements, while minimizing the need for expensive retraining. In this paper, we explore various financial applications where federated unlearning can have a significant impact, including fraud detection, portfolio management, and credit risk modeling. By allowing targeted removal of outdated, erroneous, or sensitive data, federated unlearning enhances the agility of financial models, enabling institutions to keep pace with the dynamic financial landscape. Using practical numerical examples, we demonstrate how unlearning improves model accuracy and decision-making while maintaining data privacy across distributed systems. This paper underscores the critical role of federated unlearning in addressing the challenges of modern financial institutions, offering insights into its practical applications and future potential.

federated unlearning; graph neural network; digital asset; cryptocurrency; financial market

Business, Economics and Management, Business and Management

* All users must log in before leaving a comment