Abstract: Current state-of-the-art plug-and-play countermeasures for mitigating adversarial examples (i.e., purification and detection) exhibit several fatal limitations, impeding their deployment in ...
Harnessing Transferable Adversarial Examples via Multilayer Attention-Guided Spatial Transformations
Abstract: Transfer-based adversarial attacks are key for evaluating the robustness of deep neural networks (DNNs) in black-box settings, yet their effectiveness is often constrained by limited ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results