ASD uses multi-scale spectral decomposition via DWT integrated with adversarial training to achieve SOTA robustness against patch and texture adversarial attacks, outperforming prior defenses by 21.73% under adaptive adversaries.
Imagenet large scale visual recognition challenge,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Defending against Patch-Based and Texture-Based Adversarial Attacks with Spectral Decomposition
ASD uses multi-scale spectral decomposition via DWT integrated with adversarial training to achieve SOTA robustness against patch and texture adversarial attacks, outperforming prior defenses by 21.73% under adaptive adversaries.