Differentially Private Federated Learning: Privacy and Utility Analysis of Output Perturbation and DP-SGD

Keyphrases

• Federated Learning 100%
• Differential Privacy 100%
• Differentially Private 100%
• Learning Privacy 100%
• Utility Analysis 100%
• Output Perturbation 100%
• Output Differential 100%
• Privacy Analysis 100%
• Machine Learning Models 66%
• Training Data 33%
• Privacy Concerns 33%
• Private Data 33%
• Computational Efficiency 16%
• Success Rate 16%
• Machine Learning 16%
• Stochastic Gradient Descent 16%
• Privacy-utility Tradeoff 16%
• Communication Efficiency 16%
• Sharing Model 16%
• Mathematical Definitions 16%
• Multiple Entities 16%
• Distributed Location 16%
• Perturbation Mechanism 16%
• Membership Inference Attack 16%
• Model Privacy 16%
• Membership Disclosure 16%
• Defense Techniques 16%

Computer Science

• Federated Learning 100%
• Differential Privacy 100%
• Utility Analysis 100%
• Machine Learning 83%
• Learning System 83%
• Training Data 33%
• Gradient Descent 16%
• Sensitive Informations 16%
• Computational Efficiency 16%
• Perturbation Mechanism 16%
• Membership Inference Attack 16%