nina-he / dast

A novel data-free model stealing method based on GAN

DaST: Data-free Substitute Training for Adversarial Attacks

A novel adversarial data-free substitute training for adversarial attacks. This work has been accepted by CVPR2020 as oral presentation.

Abstract: Machine learning models are vulnerable to adversarial examples. For the black-box setting, current substitute attacks need pre-trained models to generate adversarial examples. However, pre-trained models are hard to obtain in real-world tasks. In this paper, we propose a data-free substitute training method (DaST) to obtain substitute models for adversarial black-box attacks without the requirement of any real data. To achieve this, DaST utilizes specially designed generative adversarial networks (GANs) to train the substitute models. In particular, we design a multi-branch architecture and label-control loss for the generative model to deal with the uneven distribution of synthetic samples. The substitute model is then trained by the synthetic samples generated by the generative model, which are labeled by the attacked model subsequently. The experiments demonstrate the substitute models produced by DaST can achieve competitive performance compared with the baseline models which are trained by the same train set with attacked models. Additionally, to evaluate the practicability of the proposed method on the real-world task, we attack an online machine learning model on the Microsoft Azure platform. The remote model misclassifies 98.35% of the adversarial examples crafted by our method. To the best of our knowledge, we are the first to train a substitute model for adversarial attacks without any real data.

Link: https://arxiv.org/abs/2003.12703