Embedding-Based Federated Data Sharing via Differentially Private Conditional VAEs

2025-07-03

AI Summary

Federated learning (FL) in medical imaging faces challenges of data scarcity and privacy constraints, leading to high communication overhead and poor generalization on downstream tasks. Method: We propose an embedding-driven privacy-preserving federated data sharing framework. It leverages a pre-trained foundation model to extract compact image embeddings; clients collaboratively train a differentially private conditional variational autoencoder (DP-CVAE) locally to model the global data distribution efficiently. Contribution/Results: By innovatively integrating differential privacy (DP) with the CVAE architecture, our method achieves strict privacy guarantees (ε ≤ 2) while improving generation fidelity. The learned embeddings support multi-task transfer learning, requiring only 20% of the parameters of baseline models and significantly reducing communication costs. Experiments demonstrate superior performance over standard FL classifiers and DP-CGAN across classification and segmentation tasks, achieving a favorable trade-off among strong privacy preservation, high generalization capability, and low communication overhead.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Security and Privacy: Privacy-enhancing technologies
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Deep Learning (DL) has revolutionized medical imaging, yet its adoption is constrained by data scarcity and privacy regulations, limiting access to diverse datasets. Federated Learning (FL) enables decentralized training but suffers from high communication costs and is often restricted to a single downstream task, reducing flexibility. We propose a data-sharing method via Differentially Private (DP) generative models. By adopting foundation models, we extract compact, informative embeddings, reducing redundancy and lowering computational overhead. Clients collaboratively train a Differentially Private Conditional Variational Autoencoder (DP-CVAE) to model a global, privacy-aware data distribution, supporting diverse downstream tasks. Our approach, validated across multiple feature extractors, enhances privacy, scalability, and efficiency, outperforming traditional FL classifiers while ensuring differential privacy. Additionally, DP-CVAE produces higher-fidelity embeddings than DP-CGAN while requiring $5{ imes}$ fewer parameters.

Problem

• Addresses medical data scarcity and privacy constraints in deep learning
• Reduces communication costs and single-task limitations in federated learning
• Ensures privacy and efficiency via differentially private generative models

Innovation

• DP-CVAE for privacy-aware data sharing
• Foundation models for compact embeddings
• Reduces parameters and enhances fidelity