• Version
• Download 27
• File Size 276.45 KB
• File Count 1
• Create Date 7 October 2025
• Last Updated 7 October 2025

Privacy-preserving data mining and federated learning methods

 

 

Vidya Gadhave1, Vaishnavi Deshkmukh2

1Ms.Vidya Gadhave, Computer Science Department & Dr. D. Y. Patil Arts, Commerce, Science College, Pimpri

2Ms.Vaishnavi Deshkmukh, Computer Science Department & Dr. D. Y. Patil Arts, Commerce, Science College, Pimpri

 

 

Abstract

Privacy-preserving federated learning (PPFL) represents a paradigmatic shift in collaborative machine learning, addressing critical privacy concerns while enabling distributed model training across multiple organizations without compromising sensitive data. This research presents a comprehensive analysis of privacy-enhancing techniques integrated with federated learning frameworks, demonstrating how differential privacy, homomorphic encryption, and secure multi-party computation can provide robust privacy guarantees while maintaining model utility. Through systematic evaluation across healthcare, finance, and IoT applications, our findings reveal that PPFL can achieve up to 94% model accuracy while reducing privacy risks by over 60% compared to centralized approaches. The study evaluates trade-offs between privacy guarantees, communication overhead, and computational efficiency, showing that hybrid approaches combining multiple privacy techniques offer optimal performance with privacy budgets as low as ε=0.1 for differential privacy implementations. These results demonstrate the practical viability of deploying privacy-preserving federated learning systems in real-world scenarios where data sensitivity and regulatory compliance are paramount

Download