Guardians of Image Quality: Benchmarking Defenses Against Adversarial Attacks on Image-Quality Metrics

Aleksandr Gushchin, Khaled Abud, Georgii Bychkov, Ekaterina Shumitskaya, Anna Chistyakova, Sergey Lavrushkin, Bader Rasheed, Kirill Malyshev, Dmitriy S. Vatolin, Anastasia Antsiferova

Keywords: adversarial defenses, image quality assessment, adversarial attacks, image quality metrics, benchmark

TL;DR: This paper presents a new benchmark of defense methods against adversarial attacks on image quality metrics

Abstract: Most modern image-quality-assessment (IQA) metrics are based on neural networks, which makes the adversarial robustness of these metrics a critical concern. This paper presents the first comprehensive study of IQA defense mechanisms in response to adversarial attacks on these metrics. We systematically evaluated 29 defense strategies - including adversarial purification, adversarial training, and certified robustness - and applied 14 adversarial attack algorithms in both adaptive and nonadaptive settings to compare these defenses on nine no-reference IQA metrics. Our analysis of the differences between defenses and their applicability to IQA metrics recognizes that a defense technique should preserve IQA scores and image quality. Our proposed benchmark aims to guide the development of IQA defense methods and can evaluate new methods; the latest results are at link hidden for blind review.

Primary Area: datasets and benchmarks

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 2926