MAD: Meta Adversarial Defense Benchmark

X. Peng, D. Zhou, G. Sun +2

Adversarial training (AT) is a prominent technique employed by deep learning models to defend against adversarial attacks, and to some extent, enhance model robustness. However, th…