Siamese Denoising Autoencoders for Enhancing Adversarial Robustness in Medical Image Analysis

Deep learning models have achieved groundbreaking results in computer vision; however, their vulnerability to adversarial examples persists. Adversarial examples, generated by adding minute perturbations to images, lead to misclassification and pose serious threats to real-world applications of deep...

Full description

Saved in:
Bibliographic Details
Main Authors: Jaesung Shim, Kyuri Jo
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11005974/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849714388919910400
author Jaesung Shim
Kyuri Jo
author_facet Jaesung Shim
Kyuri Jo
author_sort Jaesung Shim
collection DOAJ
description Deep learning models have achieved groundbreaking results in computer vision; however, their vulnerability to adversarial examples persists. Adversarial examples, generated by adding minute perturbations to images, lead to misclassification and pose serious threats to real-world applications of deep learning models. This paper proposes a simple, powerful, and efficient adversarial defense method: a Siamese network-based Denoising Autoencoder (Siamese-DAE). This method addresses the reduction in classification accuracy caused by the denoising process. Experiments on Chest X-ray, Brain MRI, Retina, and Skin images, using FGSM, PGD, DeepFool, CW, SPSA, and AutoAttack adversarial algorithms, demonstrate that the Siamese-DAE, trained to remove noise, effectively eliminates perturbations, leading to improved classification accuracy compared not only to the standard classification model but also to relevant denoising defense models.
format Article
id doaj-art-d1728c8ede5b409d99fd3fd94f3d6b9b
institution DOAJ
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-d1728c8ede5b409d99fd3fd94f3d6b9b2025-08-20T03:13:43ZengIEEEIEEE Access2169-35362025-01-0113863338634310.1109/ACCESS.2025.357089811005974Siamese Denoising Autoencoders for Enhancing Adversarial Robustness in Medical Image AnalysisJaesung Shim0https://orcid.org/0009-0009-1464-8127Kyuri Jo1https://orcid.org/0000-0001-9222-6346Department of Computer Engineering, Chungbuk National University, Cheongju, Republic of KoreaDepartment of Computer Engineering, Chungbuk National University, Cheongju, Republic of KoreaDeep learning models have achieved groundbreaking results in computer vision; however, their vulnerability to adversarial examples persists. Adversarial examples, generated by adding minute perturbations to images, lead to misclassification and pose serious threats to real-world applications of deep learning models. This paper proposes a simple, powerful, and efficient adversarial defense method: a Siamese network-based Denoising Autoencoder (Siamese-DAE). This method addresses the reduction in classification accuracy caused by the denoising process. Experiments on Chest X-ray, Brain MRI, Retina, and Skin images, using FGSM, PGD, DeepFool, CW, SPSA, and AutoAttack adversarial algorithms, demonstrate that the Siamese-DAE, trained to remove noise, effectively eliminates perturbations, leading to improved classification accuracy compared not only to the standard classification model but also to relevant denoising defense models.https://ieeexplore.ieee.org/document/11005974/Adversarial examplesdenoising autoencoderSiamese networkdeep learning
spellingShingle Jaesung Shim
Kyuri Jo
Siamese Denoising Autoencoders for Enhancing Adversarial Robustness in Medical Image Analysis
IEEE Access
Adversarial examples
denoising autoencoder
Siamese network
deep learning
title Siamese Denoising Autoencoders for Enhancing Adversarial Robustness in Medical Image Analysis
title_full Siamese Denoising Autoencoders for Enhancing Adversarial Robustness in Medical Image Analysis
title_fullStr Siamese Denoising Autoencoders for Enhancing Adversarial Robustness in Medical Image Analysis
title_full_unstemmed Siamese Denoising Autoencoders for Enhancing Adversarial Robustness in Medical Image Analysis
title_short Siamese Denoising Autoencoders for Enhancing Adversarial Robustness in Medical Image Analysis
title_sort siamese denoising autoencoders for enhancing adversarial robustness in medical image analysis
topic Adversarial examples
denoising autoencoder
Siamese network
deep learning
url https://ieeexplore.ieee.org/document/11005974/
work_keys_str_mv AT jaesungshim siamesedenoisingautoencodersforenhancingadversarialrobustnessinmedicalimageanalysis
AT kyurijo siamesedenoisingautoencodersforenhancingadversarialrobustnessinmedicalimageanalysis