Privacy-Preserving Ranking-Based Model Integration for Federated Learning

Abstract

Federated learning enables collaborative model training while preserving data privacy. However, existing schemes often require access to clients' dataset sizes, posing privacy risks. This study proposes a novel scheme that conceals dataset sizes using Secure Multi-Party Computation (SMPC) to securely compare and rank clients based on their dataset sizes. The rankings are utilized for model aggregation, employing ranking-based weighting to achieve accurate model reinforcement while managing computation and communication overhead. Experiments on CIFAR-10 demonstrated the proposed scheme achieves 86% accuracy, surpassing FedAvg's 85%. Additionally, we confirmed that our scheme achieved scalability and efficiency in large-scale federated learning scenarios. © 2025 The Author(s). IEEJ Transactions on Electrical and Electronic Engineering published by Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.