AdvGAN++ : Harnessing latent layers for adversary generation

Abstract

Adversarial examples are fabricated examples, indistinguishable from the original image that mislead neural networks and drastically lower their performance. Recently proposed AdvGAN, a GAN based approach, takes input image as a prior for generating adversaries to target a model. In this work, we show how latent features can serve as better priors than input images for adversary generation by proposing AdvGAN++, a version of AdvGAN that achieves higher attack rates than AdvGAN and at the same time generates perceptually realistic images on MNIST and CIFAR-10 datasets.

arXiv commentAccepted at Neural Architects Workshop, ICCV 2019

References in corpus (6)

• Conditional Generative Adversarial Nets
• Deep Residual Learning for Image Recognition
• Ensemble Adversarial Training: Attacks and Defenses
• Adversarial Attacks on Neural Network Policies
• Harnessing the Vulnerability of Latent Layers in Adversarially Trained Models
• Adaptive Adversarial Attack on Scene Text Recognition

Cited by in corpus (2)

• Using a GAN to Generate Adversarial Examples to Facial Image Recognition
• Generating Out of Distribution Adversarial Attack using Latent Space Poisoning