Federated Learning for Privacy-Preserving Financial Data Sharing

Authors: Ajay Benadict Antony Raju

DOI: https://doi.org/10.5281/zenodo.14209123

Short DOI: https://doi.org/g8rrgg

Country: USA

Full-text Research PDF File:   View   |   Download

Abstract: With the growth of digitalization, most of the financial institutions are focused on the best use of information intelligence for better strategic planning and performance improvement. However, due to the confidential content of financal information, or data, privacy is a major issue of concern and hence effective measures to protect the information while trying to use it should be developed. Thus, Federated Learning (FL) as the approach to share information between the institutions without disclosing sensitive data appears to be quite helpful to overcome these challenges.
Federated Learning actually helps in group learning but at the same time, no data is shared with each other. FL also differs from the centralised data storage where a giant financial database is created, but FL enables different institutions train a common model locally on their data and only the gradients or updates of the model are shared among the participating institutions. This approach helps manage risks of the unauthorized access on the data as well as assists in the use of multiple sources of financial data for analysis.
This paper focuses on Federated Learning as the method of implementing data sharing in the financial industry with a view of examining how this learning process can transform privacy preservation practices. We address the fundamental concepts of FL and its application in financial environments as well as the advantages, such as increased protection and data privacy compliance. Furthermore, we discuss issues including model accuracy, communication complexity, and the requirement for safe three-way aggregation. With FL, it is possible to incorporate it into the frameworks involved in sharing of financial data, institutions can strengthen their analytical processes incrementally from other institutions while observing the best practices in data privacy.
Based on our work, Federated Learning appears as a promising improvement in handling data utility while being more sensitive to privacy issues, which will eventually lead to even more efficient and secure methods of sharing financial data.

Keywords: Federated Learning, Privacy-preserving, Financial Data, Data Sharing, Machine Learning, Data Security

Paper Id: 231654

Published On: 2024-06-05

Published In: Volume 12, Issue 3, May-June 2024

Cite This: Federated Learning for Privacy-Preserving Financial Data Sharing - Ajay Benadict Antony Raju - IJIRMPS Volume 12, Issue 3, May-June 2024. DOI 10.5281/zenodo.14209123

Share this