AT GAN An Adversarial Generative Model for Non constrained Adversarial Examples

With rapid development of adversarial machine learning, numerous adversarial attack methods have been proposed . Typical attacks are based on a search in the neighborhood of input image to generate a perturbed adversarial example . In this work, we propose AT-GAN (Adversarial Transfer on Generative Adversarial Net) to train an adversarial generative model that can directly produce adversarial examples . We aim to learn the distribution of . adversarial . examples so as to generate semantically meaningful adversaries . We hope to efficiently generate diverse . diverse . adversaries that are realistic to human perception, and yields higher attack success rates against . adversarially trained models . AT-ANAT-GAN could generate non-constrained adversaries examples for any input noise, denoted as non-consistained adversarial Examples.  non-Constrained adversarial examples, denoting non-comprehensible anagrams and more effective attacks. The research was published on Tuesday, October 1, at MIT Press Press Press Conference: MIT Press Conference.

Links: PDF - Abstract

Code :


Keywords : adversarial - examples - press - generate - generative -

Leave a Reply

Your email address will not be published. Required fields are marked *