Adversarial purification for no-reference image-quality metrics: applicability study and new methods

TL;DR

This paper investigates the transferability of adversarial purification defenses from image classifiers to image quality assessment (IQA) models, evaluating various attack and defense methods to improve robustness of IQA metrics.

Contribution

It introduces a comprehensive study on applying and assessing adversarial purification techniques for IQA models, including new evaluation methods for defense efficacy.

Findings

Purification techniques can mitigate adversarial attacks on IQA models.

Transferability of defenses from classifiers to IQA is feasible.

Proposed evaluation methods effectively measure defense success.

Abstract

Recently, the area of adversarial attacks on image quality metrics has begun to be explored, whereas the area of defences remains under-researched. In this study, we aim to cover that case and check the transferability of adversarial purification defences from image classifiers to IQA methods. In this paper, we apply several widespread attacks on IQA models and examine the success of the defences against them. The purification methodologies covered different preprocessing techniques, including geometrical transformations, compression, denoising, and modern neural network-based methods. Also, we address the challenge of assessing the efficacy of a defensive methodology by proposing ways to estimate output visual quality and the success of neutralizing attacks. Defences were tested against attack on three IQA metrics -- Linearity, MetaIQA and SPAQ. The code for attacks and defences is available at: (link is hidden for a blind review).