Abstract: The transferability of adversarial examples is pivotal in black-box attacks on deep learning models. The existing transfer-based attacks typically rely on a single data augmentation ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback