Bppfl: a blockchain-based framework for privacy-preserving federated learning

Muhammad Asad, Zayed University
Safa Otoum, Zayed University

Document Type

Article

Source of Publication

Cluster Computing

Publication Date

11-26-2024

Abstract

Federated Learning (FL) offers a collaborative approach to training machine learning models while preserving data privacy. However, FL faces significant privacy and security challenges, such as identity disclosure and model inference attacks. To this end, we propose a novel Blockchain-Based Framework for Privacy-Preserving Federated Learning (BPPFL), which integrates threshold signature authentication and threshold Paillier encryption with blockchain technology. The BPPFL framework secures participant authentication and protects against internal and external threats, while the blockchain provides an immutable ledger for recording transactions and model updates, ensuring transparency and security. Experimental results show that our framework significantly reduces computation and communication overhead compared to existing methods while maintaining high model accuracy and robust privacy guarantees. Our framework enhances the security and trustworthiness of FL applications, making it suitable for domains like healthcare, finance, and the IoT.

ISSN

1386-7543

Publisher

Springer Nature

Volume

28

Issue

2

First Page

126

Last Page

126

Disciplines

Computer Sciences

Keywords

Federated learning, Blockchain, Privacy-preserving, Pallier encryption

Indexed in Scopus

no

Open Access

no

Link to Full Text

Share

COinS