Abstract: In recent years, adversarial attack methods have been deceived rather easily on deep neural networks (DNNs). In practice, adversarial patches cause misclassification that can be extremely ...