DiffAttack: Evasion Attacks Against Diffusion-Based Adversarial Purification

Neural Information Processing Systems 

Recent studies show that even advanced attacks cannot break such defenses effectively, since the purification process induces an extremely deep computational graph which poses the potential problem of vanishing/exploding gradient, high memory cost, and unbounded randomness.

Similar Docs  Excel Report  more

TitleSimilaritySource
None found