Efficient Data Collaboration Using Multi-Party Privacy Preserving Machine Learning Framework

Abstract

In a modern era where data-driven insights are the foundation of technological advancements preserving the privacy and security of sensitive information while harnessing the collective intelligence of multiple parties is imperative. This research presents a Secure Collaborative Learning Algorithm (SCLA) designed to facilitate efficient multi-party machine learning without compromising data privacy. Our research focus is on leveraging existing, secure databases without requiring an additional data collection process. SCLA seamlessly integrates homomorphic encryption and Federated Learning (FL) to enable secure data collaboration among various stakeholders. The proposed algorithm aggregates model updates in a privacy-preserving manner, demonstrating enhanced model accuracy, competitive convergence speed, and robust scalability. By carefully balancing privacy preservation and learning efficiency, the SCLA showcases a promising avenue for privacy-focused collaborative learning using existing data repositories.