A Tale of Evil Twins: Adversarial Inputs versus Poisoned Models

Keyphrases

• Attack Vectors 100%
• Evil Twin 100%
• Adversarial Inputs 100%
• Vulnerability 25%
• Detection Method 25%
• Adversary 25%
• Unified Framework 25%
• Mutual Reinforcement 25%
• Dynamic Interaction 25%
• Reinforcement Effect 25%
• Attack Model 25%
• Lack of Understanding 25%
• Deep Neural Network 25%
• Realistic Settings 25%
• Persuasiveness 25%
• Deep Learning System 25%
• Design Spectra 25%
• Backdoor Attack 25%

Computer Science

• Deep Neural Network 100%
• Potential Countermeasure 66%
• Neural Network Model 33%
• Research Direction 33%
• Unified Framework 33%
• Attack Model 33%
• Detection Method 33%
• Technical Challenge 33%
• Dynamic Interaction 33%
• Learning System 33%
• Deep Learning Method 33%
• Optical Spectrum 33%
• Backdoor Attack 33%
• Empirical evidence 33%

Mathematics

• Deep Neural Network 100%
• Network Model 33%
• Detection Method 33%
• Deep Learning Method 33%
• Empirical evidence 33%