Attack Type Agnostic Perceptual Enhancement of Adversarial Images

TL;DR

This paper introduces an attack type agnostic method to enhance the perceptual quality of adversarial images, reducing noise and artificial colors while maintaining attack effectiveness, thereby improving human interpretability.

Contribution

It presents a novel perceptual enhancement technique applicable to various adversarial attacks, improving image quality without compromising attack success.

Findings

Lower Euclidean distance values in enhanced images

Distance reduction averages 22% across attacks and networks

Maintains adversarial attack performance after enhancement

Abstract

Adversarial images are samples that are intentionally modified to deceive machine learning systems. They are widely used in applications such as CAPTHAs to help distinguish legitimate human users from bots. However, the noise introduced during the adversarial image generation process degrades the perceptual quality and introduces artificial colours; making it also difficult for humans to classify images and recognise objects. In this letter, we propose a method to enhance the perceptual quality of these adversarial images. The proposed method is attack type agnostic and could be used in association with the existing attacks in the literature. Our experiments show that the generated adversarial images have lower Euclidean distance values while maintaining the same adversarial attack performance. Distances are reduced by 5.88% to 41.27% with an average reduction of 22% over the different attack and network types.