A Comparative Analysis of Image Steganalysis Methods: Handcrafted, Deep Learning, and Hybrid Paradigms
Journal of Al-Farabi for Engineering Sciences · 2025
DOI: 10.59746/3tcsyt78 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Al-Farabi for Engineering Sciences, 2025. Verfasst von Saif Salahaldin Affat und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Over the last decade, image steganalysis, the science of detecting hidden information in digital images, has undergone a fundamental transformation, from manually engineered statistical feature methods to sophisticated deep learning architectures that can automatically extract features. This survey presents a systematic and detailed review of image steganalysis techniques, which can be grouped into four main categories: 1) Traditional intensity and texture-based methods including threshold-based methods and Grey Level Co-occurrence Matrix (GLCM) features; 2) Handcrafted feature-based methods such as the Spatial Rich Model (SRM), SPAM, DCTR, and GFR with ensemble classifiers; 3) Machine learning and deep learning methods including CNN-based architectures such as Xu-Net, Ye-Net, Yedroudj-Net, and SRNet; and 4) Hybrid fusion-based frameworks combining handcrafted and learned features to enhance detection robustness. The survey also includes common evaluation criteria, popular benchmark datasets such as BOSSbase, ALASKA II and IStego100K, and a comparison of stated detection accuracies in terms of methodologies and datasets. Key findings indicate that deep learning approaches consistently outperform traditional methods, with reported accuracies of up to 99% on some benchmarks, but important challenges remain unaddressed, such as the cover-source mismatch, low-payload detection, and the generalization gap between controlled laboratory conditions and real-world social media transmission. This review highlights these outstanding difficulties and discusses the future research directions that need to be followed in order to produce universal, resilient and computationally efficient steganalysis systems that can work efficiently in realistic deployment situations.
Zitieren
APA
Affat, S. S., & Ahmed, I. T. (2025). A Comparative Analysis of Image Steganalysis Methods: Handcrafted, Deep Learning, and Hybrid Paradigms. Journal of Al-Farabi for Engineering Sciences. https://doi.org/10.59746/3tcsyt78
Vancouver
Affat SS, Ahmed IT. A Comparative Analysis of Image Steganalysis Methods: Handcrafted, Deep Learning, and Hybrid Paradigms. Journal of Al-Farabi for Engineering Sciences. 2025. doi:10.59746/3tcsyt78
BibTeX
@article{Affat2025,
author = {Affat, Saif Salahaldin and Ahmed, Ismael Taha},
title = {A Comparative Analysis of Image Steganalysis Methods: Handcrafted, Deep Learning, and Hybrid Paradigms},
journal = {Journal of Al-Farabi for Engineering Sciences},
year = {2025},
doi = {10.59746/3tcsyt78}
}
EUVIMED