No video available

Please refer to the materials section for this topic.

Threat Models: White-Box, Black-Box, and Gray-Box Security Assumptions

Learning Objectives

•Attacker capabilities: gradient access, query budgets, confidence scores
•Norm-constrained perturbations (L0, L2, L-infinity)
•Targeted vs Untargeted misclassification goals

Weekly Outcome

Master the core concepts of adversarial threat models & white-box attacks.

Resources

Adversarial Machine Learning: Attacks, Defenses, and Open ...Adversarial Machine Learning Framed by Threat Models, Attack ...White, Gray & Black-Box Attacks Overview