Black-box adversarial attacks against image quality assessment models

Published: 01 Jan 2025, Last Modified: 11 Apr 2025Expert Syst. Appl. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•A new adversarial attack problem on image quality assessment (IQA) models.•A new loss function tailored for the proposed attack problem.•An efficient and effective adversarial example generation algorithm.•Results show that the IQA models are vulnerable to the proposed attacks.•Pointing out a new research direction in IQA community.
Loading