On Frequency Domain Adversarial Vulnerabilities of Volumetric Medical Image Segmentation
Hanif, Asif Naseer, Muzammal Khan, Salman Khan, Fahad Shahbaz
Hanif, Asif
Naseer, Muzammal
Khan, Salman
Khan, Fahad Shahbaz
Supervisor
Department
Computer Vision
Embargo End Date
Type
Conference proceeding
Date
2025
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
In safety-critical domains like healthcare, resilience of deep learning models towards adversarial attacks is crucial. Volumetric medical image segmentation is a fundamental task, providing critical insights for diagnosis. This paper introduces a novel frequency domain adversarial attack targeting 3D medical data, revealing vulnerabilities in segmentation models. By manipulating the frequency spectrum (low, middle, and high bands), we assess its impact on model performance. Unlike pixel-based 2D attacks, our method continuously perturbs 3D samples with minimal information loss, achieving high fooling rates at lower computational costs and superior black-box transferability, while maintaining perceptual quality. Our code is publicly available at Github††https://github.com/asif-hanif/saa.
Citation
A. Hanif, M. Naseer, S. Khan and F. S. Khan, "On Frequency Domain Adversarial Vulnerabilities of Volumetric Medical Image Segmentation," 2025 IEEE 22nd International Symposium on Biomedical Imaging (ISBI), Houston, TX, USA, 2025, pp. 01-05, doi: 10.1109/ISBI60581.2025.10981075
Source
2025 IEEE 22nd International Symposium on Biomedical Imaging (ISBI)
Conference
IEEE International Symposium on Biomedical Imaging, 2025
Keywords
Medical image segmentation, Adversarial attack, Spectral domain attack, Attack transferability
Subjects
Source
IEEE International Symposium on Biomedical Imaging, 2025
Publisher
IEEE