DAPAS : Denoising Autoencoder to Prevent Adversarial attack in Semantic Segmentation

TL;DR

This paper proposes using a denoising autoencoder to defend semantic segmentation models against adversarial attacks by removing perturbations and restoring original images, enhancing model robustness.

Contribution

It introduces a novel defense method employing denoising autoencoders specifically for semantic segmentation to counter adversarial perturbations.

Findings

Denoising autoencoders effectively reduce adversarial noise in semantic segmentation.

The method improves model robustness against various noise distributions.

Experimental results show increased accuracy in adversarial settings.

Abstract

Nowadays, Deep learning techniques show dramatic performance on computer vision area, and they even outperform human. But it is also vulnerable to some small perturbation called an adversarial attack. This is a problem combined with the safety of artificial intelligence, which has recently been studied a lot. These attacks have shown that they can fool models of image classification, semantic segmentation, and object detection. We point out this attack can be protected by denoise autoencoder, which is used for denoising the perturbation and restoring the original images. We experiment with various noise distributions and verify the effect of denoise autoencoder against adversarial attack in semantic segmentation.