Privacy-Preserving Federated Data Governance Framework for Secure Parameter Exchange in Distance Education

Abstract

This study presents a hybrid privacy-preserving framework for secure parameter exchange in federated learning (FL) within distance education. The framework integrates Differential Privacy, quantum-resistant NTRU Encryption, and Blockchain technology to ensure data regulatory compliance. Logistic Regression was employed as the machine learning model in the federated learning scenario using real-world educational datasets. Performance evaluation in terms of model performance over federated rounds, latency and communication overhead per training round, noise effect on accuracy, and trust and audit traceability demonstrates the effectiveness of the novel tri-layered governance framework for secure parameter exchange uniquely adapted for distance education settings. Compared to a baseline FL model using Differential Privacy and Secure Aggregation, the proposed method achieves higher accuracy (<tex>$\mathbf{0. 8 7}$</tex> vs. <tex>$\mathbf{0. 8 1}$</tex>), lower latency (0.2 s vs. 0.4s), and reduced communication overhead (0.55 s vs. 0.85 s) per round.