On the detection of adversarial attacks against deep neural networks

Mathematics

• Image Recognition 100%
• Attack 93%
• Neural Networks 86%
• Learning 55%
• Human 55%
• Credibility 39%
• Prediction Model 39%
• Sample space 37%
• Ill-posedness 37%
• Pattern Recognition 34%
• Traffic 33%
• Architecture 33%
• Machine Learning 33%
• High Accuracy 31%
• Small Perturbations 30%
• False 30%
• Robustness 25%
• High-dimensional 25%
• Prediction 24%
• Output 22%
• Perturbation 21%
• Model 20%

Engineering & Materials Science

• Deep neural networks 89%
• Deep learning 54%
• Traffic signs 40%
• Pattern recognition 26%
• Labels 25%
• Machine learning 22%
• Feedback 19%