Towards Robust ML: Defending Against Adversarial Attacks
Summary
Towards robust machine learning: defending against adversarial attacks. DIPPAD: Denoising Diffusion-based Adversarial Patch Decontamination. Methods for adversarial training, certified defenses, and adversarial purification.
Themes
Keywords
adversarial attacks, robust ML, DIPPAD, diffusion models, adversarial defense
Poster
Click image to open full size