Generating Adversarial Examples With Conditional Generative Adversarial Net

TL;DR

This paper introduces two novel conditional generative adversarial network models that efficiently generate robust adversarial examples, significantly reducing generation time compared to traditional methods like FGSM.

Contribution

The paper proposes new generative models for adversarial example creation that are faster and more robust than existing techniques, with a distinctive training strategy.

Findings

Models reduce attack generation time by 80%

Enhanced robustness of generated adversarial examples

Outperforms traditional methods like FGSM in efficiency

Abstract

Recently, deep neural networks have significant progress and successful application in various fields, but they are found vulnerable to attack instances, e.g., adversarial examples. State-of-art attack methods can generate attack images by adding small perturbation to the source image. These attack images can fool the classifier but have little impact to human. Therefore, such attack instances are difficult to generate by searching the feature space. How to design an effective and robust generating method has become a spotlight. Inspired by adversarial examples, we propose two novel generative models to produce adaptive attack instances directly, in which conditional generative adversarial network is adopted and distinctive strategy is designed for training. Compared with the common method, such as Fast Gradient Sign Method, our models can reduce the generating cost and improve robustness and has about one fifth running time for producing attack instance.