Privacy-Preserving Generative Models: A Comprehensive Survey

2025-02-05

AI Summary

This work addresses the dual challenges of privacy leakage and utility degradation in generative models (GANs and VAEs). Methodologically, it conducts a systematic review of 100 papers and proposes— for the first time—the Privacy-Utility Co-Taxonomy for generative models, a two-dimensional classification framework spanning model architecture, training mechanisms, and data granularity. It further introduces unified, multi-level evaluation metrics and establishes the first comprehensive assessment framework compatible with both GANs and VAEs. Through bibliometric analysis, cross-model comparison, and privacy threat modeling, the study identifies critical technical bottlenecks and synthesizes six key future research directions. The contributions include: (1) the first domain-specific taxonomy bridging privacy and utility in generative modeling; (2) a standardized, model-agnostic evaluation framework; and (3) a rigorous, evidence-based roadmap addressing the longstanding gap in systematic survey and taxonomic modeling for privacy-preserving generative AI.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Privacy
• Philosophy and Ethics of AI: Privacy & Security

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Security and Privacy: Security and privacy of machine learning and AI applications

Abstract

Despite the generative model's groundbreaking success, the need to study its implications for privacy and utility becomes more urgent. Although many studies have demonstrated the privacy threats brought by GANs, no existing survey has systematically categorized the privacy and utility perspectives of GANs and VAEs. In this article, we comprehensively study privacy-preserving generative models, articulating the novel taxonomies for both privacy and utility metrics by analyzing 100 research publications. Finally, we discuss the current challenges and future research directions that help new researchers gain insight into the underlying concepts.

Problem

• Systematically categorize privacy and utility in GANs and VAEs
• Develop taxonomies for privacy-preserving generative models
• Discuss challenges and future research directions in privacy preservation

Innovation

• privacy-preserving generative models
• novel taxonomies metrics
• analyzing 100 publications