Defense against Adversarial Attacks Using High-Level Representation Guided Denoiser

TL;DR

This paper introduces a high-level representation guided denoiser (HGD) that effectively defends neural networks against adversarial attacks by reducing error amplification, improving robustness, and generalizing across models and unseen classes.

Contribution

The paper proposes HGD, a novel denoising method that overcomes error amplification and enhances adversarial robustness, outperforming existing ensemble adversarial training methods.

Findings

HGD improves robustness to white-box and black-box attacks.

HGD generalizes well to unseen images and classes.

HGD won first place in the NIPS defense competition.

Abstract

Neural networks are vulnerable to adversarial examples, which poses a threat to their application in security sensitive systems. We propose high-level representation guided denoiser (HGD) as a defense for image classification. Standard denoiser suffers from the error amplification effect, in which small residual adversarial noise is progressively amplified and leads to wrong classifications. HGD overcomes this problem by using a loss function defined as the difference between the target model's outputs activated by the clean image and denoised image. Compared with ensemble adversarial training which is the state-of-the-art defending method on large images, HGD has three advantages. First, with HGD as a defense, the target model is more robust to either white-box or black-box adversarial attacks. Second, HGD can be trained on a small subset of the images and generalizes well to other images and unseen classes. Third, HGD can be transferred to defend models other than the one guiding it. In NIPS competition on defense against adversarial attacks, our HGD solution won the first place and outperformed other models by a large margin.