Discriminator-Free Generative Adversarial Attack

TL;DR

This paper introduces SSAE, a discriminator-free generative adversarial attack method that efficiently creates high-quality adversarial examples by focusing on salient regions, leading to effective model collapse.

Contribution

It proposes a novel discriminator-free approach using saliency maps and feature disentanglement for generating adversarial examples, overcoming GAN training difficulties.

Findings

SSAE effectively collapses target models across various datasets.

Generated adversarial examples have high visual quality.

The method reduces training complexity by removing the discriminator.

Abstract

The Deep Neural Networks are vulnerable toadversarial exam-ples(Figure 1), making the DNNs-based systems collapsed byadding the inconspicuous perturbations to the images. Most of the existing works for adversarial attack are gradient-based and suf-fer from the latency efficiencies and the load on GPU memory. Thegenerative-based adversarial attacks can get rid of this limitation,and some relative works propose the approaches based on GAN.However, suffering from the difficulty of the convergence of train-ing a GAN, the adversarial examples have either bad attack abilityor bad visual quality. In this work, we find that the discriminatorcould be not necessary for generative-based adversarial attack, andpropose theSymmetric Saliency-based Auto-Encoder (SSAE)to generate the perturbations, which is composed of the saliencymap module and the angle-norm disentanglement of the featuresmodule. The advantage of our proposed method lies in that it is notdepending on discriminator, and uses the generative saliency map to pay more attention to label-relevant regions. The extensive exper-iments among the various tasks, datasets, and models demonstratethat the adversarial examples generated by SSAE not only make thewidely-used models collapse, but also achieves good visual quality.The code is available at https://github.com/BravoLu/SSAE.