Comparative Analysis of Frequency-Domain and Spatial-Domain Deep Learning Methods for Detecting Diffusion-Generated Human Faces

Ziad Karim Tabbouche

Comparative Analysis of Frequency-Domain and Spatial-Domain Deep Learning Methods for Detecting Diffusion-Generated Human Faces

Author(s)	Ziad Karim Tabbouche
Country	India
Abstract	The rapid advancement of diffusion-based image generation models has significantly improved the realism of synthetic human faces, creating major challenges for digital media authentication and forensic analysis. Traditional deepfake detection approaches primarily rely on spatial-domain artifacts, but modern diffusion models generate highly realistic images that minimize visible inconsistencies. Consequently, frequency-domain analysis has emerged as an alternative strategy for identifying subtle spectral anomalies left during the image synthesis process. This paper presents a comparative analysis of frequency-domain and spatial-domain deep learning methods for detecting diffusion-generated human faces. The study evaluates representative convolutional neural networks, transformer-based architectures, and hybrid dual-domain models across publicly available datasets. Experimental findings demonstrate that spatial-domain methods achieve strong performance on known datasets but suffer from reduced generalization when confronted with unseen diffusion models. Frequency-domain approaches exhibit superior robustness against compression and cross-dataset variations due to their ability to capture hidden spectral artifacts. Furthermore, hybrid models integrating both domains outperform single-domain approaches by leveraging complementary discriminative features. The paper also discusses challenges related to dataset bias, adversarial robustness, and model interpretability. The results suggest that future detection systems should adopt collaborative spatial-frequency learning frameworks to improve reliability against increasingly sophisticated generative models.
Keywords	Diffusion Models, Deepfake Detection, Frequency Domain Analysis, Spatial Domain Learning, CNN, Vision Transformers, Synthetic Face Detection, AI-Generated Images, Digital Forensics, Multimedia Security
Published In	Volume 8, Issue 4, July-August 2026
Published On	2026-08-10

View / Download PDF File

About IJFMR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijfmr.com

International Journal For Multidisciplinary Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Comparative Analysis of Frequency-Domain and Spatial-Domain Deep Learning Methods for Detecting Diffusion-Generated Human Faces

Share this