Rob-GAN: Generator, Discriminator, and Adversarial Attacker

Abstract

We study two important concepts in adversarial deep learning-adversarial training and generative adversarial network (GAN). Adversarial training is the technique used to improve the robustness of discriminator by combining adversarial attacker and discriminator in the training phase. GAN is commonly used for image generation by jointly optimizing discriminator and generator. We show these two concepts are indeed closely related and can be used to strengthen each other-adding a generator to the adversarial training procedure can improve the robustness of discriminators, and adding an adversarial attack to GAN training can improve the convergence speed and lead to better generators. Combining these two insights, we develop a framework called Rob-GAN to jointly optimize generator and discriminator in the presence of adversarial attacksthe generator generates fake images to fool discriminator; the adversarial attacker perturbs real images to fool discriminator, and the discriminator wants to minimize loss under fake and adversarial images. Through this end-to-end training procedure, we are able to simultaneously improve the convergence speed of GAN training, the quality of synthetic images, and the robustness of discriminator under strong adversarial attacks. Experimental results demonstrate that the obtained classifier is more robust than state-of-the-art adversarial training approach [23], and the generator outperforms SN-GAN on ImageNet-143.

Deep Generative Modelling: A Comparative Review of VAEs, GANs, Normalizing Flows, Energy-Based and Autoregressive Models

“…While strong discriminator regularisation stabilises training, it allows the generator to make small changes and trick the discriminator, making convergence very slow. Rob-GAN [141], include an adversarial attack step [149] that perturbs real images to trick the discriminator without altering the content inordinately, adapting the GAN objective into a min-max-min problem. This provides a weaker regularisation, enforcing small Lipschitz values locally rather than globally.…”

Noise Homogenization via Multi-Channel Wavelet Filtering for High-Fidelity Sample Generation in Gans

“…Some of the current methods merely focused on discriminator learning, while the others were more specific on the generator, or both [9]. That being said, there are a lot of complicated models proposed by stacking multiple architectures [16,17]. Among these areas, there is another trend specific on the signal conditions, i.e., conditional training with labels [10] and unsupervised GANs trained without prior labels [9].…”

Detect and defense against adversarial examples in deep learning using natural scene statistics and adaptive denoising

“…The adversarial training techniques consist in including AEs at the training stage of the model to build a robust classifier. Authors in [9,34,35] used benign samples with adversarial samples as data augmentation in the training process. In practice, different attacks can be used to generate the AEs.…”