Privacy-Enhancing Machine Learning with Differential Privacy

Read full text Summarise with VIRAT
Authors: Stephen Eteng
Cited by: 0
References: 55

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Privacy concerns in machine learning have become increasingly critical as AI systems process sensitive personal, financial, healthcare, and behavioral data. Traditional machine learning approaches often require centralizing large datasets, raising risks of data breaches, unauthorized access, and privacy violations. Differential privacy (DP) has emerged as a formal framework to quantify and guarantee privacy protection while enabling model training on sensitive data. By introducing controlled noise into data or model computations, differential privacy ensures that the inclusion or exclusion of a single individual’s data has minimal impact on model outputs, providing strong mathematical privacy guarantees. Privacy-enhancing machine learning (PEML) techniques incorporating differential privacy are applied across domains such as healthcare, finance, recommendation systems, and federated learning. These approaches balance the trade-off between data utility and privacy, enabling collaborative AI without exposing sensitive information. Challenges include maintaining model accuracy under privacy constraints, computational overhead, and adaptive threat resistance. Future directions involve integrating DP with federated and decentralized learning, adaptive privacy budgets, algorithmic optimization, and explainable privacy-aware AI. Differential privacy represents a cornerstone in building trustworthy, privacy-preserving AI systems in an increasingly data-driven world.

Fields of Study & Subjects

Privacy-Preserving Technologies in Data Privacy Security and Data Protection Cloud Data Security Solutions

Publication details

Journal / source
International Journal of Multidisciplinary Sciences and Technology
Volume
2
Issue
1
Pages
28–33
Publication date
20 March 2026
Language
EN