GAN-based feature-level supportivemethod for improved adversarial attacks on face recognition

With the rapid development of deep neural networks (DNN), DNN-based face recognition technologies are also achieving great success and have been widely used in various applications requiring high-accuracy and robustness against attacks. However, deep neural networks are known to be vulnerable to adversarial attacks, performed using images added with well-designed perturbations. To enhance security of DNN-based face recognition, we need to know more about mechanisms of relate technologies. In this paper, we propose a feature-level supportive method, biasGAN, to improve the performance of universal adversarial attack methods. We insert this image to image translation preprocessor before conducting adversarial examples generation. BiasGAN will search in the potential face image space and can generate images with biased face feature characteristics, causing generated face images to be easier to perturb efficiently. Experiment results show that this approach improves both fooling ratio and average perturbation size significantly at different perturbation levels