Guessing smart: Biased sampling for efficient black-box adversarial attacks

Keyphrases

• Biased Sampling 100%
• Black-box Adversarial Attack 100%
• ImageNet 50%
• Wide Margin 50%
• Image Classification 50%
• Black Box 50%
• Class-specific 50%
• Attacker 50%
• Confidence Score 50%
• Image Frequency 50%
• Adversarial Examples 50%
• Sampling Framework 50%
• Gain Efficiency 50%
• Targeted Attack 50%
• Surrogate Gradient 50%
• Google Cloud Vision API 50%
• Decision Basis 50%
• Boundary Attack 50%

Computer Science

• Image Classification 100%
• Adversarial Machine Learning 100%
• Attackers 100%
• Confidence Score 100%
• Adversarial Example 100%
• Targeted Attack 100%
• Domain Knowledge 100%
• Image Frequency 100%