MC-GAN: An Adversarial Sample Defense Algorithm

The current adversarial sample defense algorithms are plagued by issues such as poor defense effectiveness and high training costs. In order to enhance sample classification accuracy while reducing training costs, we have introduced a novel defense algorithm based on model compression within the framework of Generative Adversarial Networks (GANs). This algorithm integrates GANs with model compression techniques, and the results of testing on public datasets have demonstrated an improved accuracy in defending against adversarial samples along with reduced training costs.