论文部分内容阅读
Deep learning model is vulnerable to adversarial examples in the task of image classification. In this paper, a cluster-based method for defending against adversarial examples is proposed. Each adversarial example before attacking a classifier is reconstr