DFT-Based Adversarial Attack Detection in MRI Brain Imaging: Enhancing Diagnostic Accuracy in Alzheimer's Case Studies

TL;DR

The paper tackles the vulnerability of CNN-based MRI brain classifiers for Alzheimer's disease to adversarial perturbations by proposing a frequency-domain defense using the $2D$-$DFT$. A CNN-based autoencoder trained on clean data learns a latent manifold and reconstructs images in the frequency domain, enabling detection via reconstruction error and manifold distance with thresholds $t_m$ and $t_{re}$. The approach is evaluated on the Kaggle OASIS MRI dataset across multiple models (Inception v3, VGG16, ResNet-50, Simple CNN) and attacks (gradient-based and frequency-domain variants such as $DWT$-$FGSM$, $DWT$-$PGD$, etc.), showing improved robustness. This work demonstrates a practical pathway to safer medical imaging pipelines by leveraging frequency-domain representations and autoencoder-based detection, with potential for further enhancement through additional features and defenses against perceptual adversaries.

Abstract

Recent advancements in deep learning, particularly in medical imaging, have significantly propelled the progress of healthcare systems. However, examining the robustness of medical images against adversarial attacks is crucial due to their real-world applications and profound impact on individuals' health. These attacks can result in misclassifications in disease diagnosis, potentially leading to severe consequences. Numerous studies have explored both the implementation of adversarial attacks on medical images and the development of defense mechanisms against these threats, highlighting the vulnerabilities of deep neural networks to such adversarial activities. In this study, we investigate adversarial attacks on images associated with Alzheimer's disease and propose a defensive method to counteract these attacks. Specifically, we examine adversarial attacks that employ frequency domain transformations on Alzheimer's disease images, along with other well-known adversarial attacks. Our approach utilizes a convolutional neural network (CNN)-based autoencoder architecture in conjunction with the two-dimensional Fourier transform of images for detection purposes. The simulation results demonstrate that our detection and defense mechanism effectively mitigates several adversarial attacks, thereby enhancing the robustness of deep neural networks against such vulnerabilities.

Figures (4)

• Figure 1: Comparison of Original, High Frequency, and Low Frequency Images. Image is from Kaggle OASIS dataset daithal_imagesoasis_2024.
• Figure 2: Comparison of Original and Attacked Alzheimer's Disease MRI Images and Their DFT. Image is from Kaggle OASIS dataset.
• Figure 3: Architecture of U-Net-Based Autoencoder
• Figure 4: Architecture of Simple CNN