BDPL: A Boundary Differentially Private Layer Against Machine Learning Model Extraction Attacks

Keyphrases

• Decision Boundary 100%
• Machine Learning Models 100%
• Differentially Private 100%
• Model Extraction Attack 100%
• Adversary 66%
• Prediction Response 66%
• Merchandising 33%
• Large Volume 33%
• Nonlinear Model 33%
• Confidence Level 33%
• Differential Property 33%
• Computational Resources 33%
• Randomized Response 33%
• One-bit 33%
• Obfuscation 33%
• Differential Privacy 33%
• Proprietary Information 33%
• Replica Model 33%
• Perturbation-iteration Algorithm 33%
• Service API 33%
• Extraction Attacks 33%
• High Utility Patterns 33%

Computer Science

• Decision Boundary 100%
• Machine Learning 100%
• Learning System 100%
• Differential Privacy 33%
• Valuable Asset 33%
• Computational Resource 33%
• Confidence Level 33%
• Nonlinear Model 33%