Orthogonal Deep Models as Defense against Black-Box Attacks

Computer Science

• Deep Learning Method 100%
• Adversarial Machine Learning 100%
• Black-Box Attack 100%
• Computer Vision Task 50%
• Art Performance 50%
• Reference Model 50%
• Attackers 50%
• Smart Surveillance 50%
• Unique Constraint 50%
• Critical Application 50%
• Internal Representation 50%
• Regularization 50%

Mathematics

• Black Box 100%
• Deep Learning Method 50%
• Regularization 25%
• Scale Model 25%
• Reference Model 25%
• Orthogonality 25%