DATA PRIVACY-AWARE MACHINE LEARNING AND FEDERATED LEARNING: A FRAMEWORK FOR DATA SECURITY

Abstract

This study presents a comprehensive systematic review and meta-analysis of 128 peer-reviewed publications on data privacy-aware machine learning (ML) and federated learning (FL), synthesizing their theoretical foundations, computational mechanisms, and ethical implications within the evolving landscape of privacy-preserving artificial intelligence. Guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework, the research integrates multidisciplinary perspectives spanning computer science, ethics, law, and digital governance to evaluate how privacy-aware methodologies and decentralized architectures collectively enhance data protection, regulatory compliance, and algorithmic accountability. The findings reveal that differential privacy, homomorphic encryption, and secure multi-party computation constitute the principal mechanisms enabling quantifiable confidentiality without significant loss of model utility. Concurrently, federated learning has emerged as a scalable and policy-aligned framework that decentralizes computation, ensuring data sovereignty and compliance with international privacy regulations such as GDPR, HIPAA, and CCPA. The meta-analysis indicates that integrated privacy-preserving federated systems achieve an average model accuracy of 93%, reduce data leakage risks by 68%, and improve overall energy efficiency by 22% relative to traditional centralized architectures. However, the study also identifies persistent challenges, including communication bottlenecks, heterogeneity in non-identically distributed datasets, trade-offs between privacy and interpretability, and the underexplored environmental costs of encryption and distributed computation. Despite these limitations, the synthesis affirms that privacy-aware federated learning represents a paradigm shift in artificial intelligence—from reactive data protection to proactive privacy-by-design computation. By uniting technical innovation, ethical governance, and policy coherence, this study establishes a holistic framework that redefines data privacy as both a computational property and a moral imperative in the era of intelligent, decentralized automation.