Manuscript received December 23, 2023; revised March 1, 2024; accepted March 13, 2024; published September 23, 2024

Abstract—This paper explores the use of deep learning algorithms in steganography detection. More specifically, it examines deep learning-based binary classification to distinguish between stego and non-stego images from the three steganography algorithms, The Wavelet Obtained Weights (WOW), Spatial Universal Wavelet Relative Distortion (S-UNIWARD), Highly Undetectable Steganography (HUGO). It also highlights the lack of research to develop a practical universal image steganography detection system using trained deep learning. The proposed farmwork combines multiple detection deep learning architectures to create a universal Deep Convolutional Neural Network (Deep-CNN). In this paper, we evaluate Deep-CNN-based image steganography detection techniques trained on images extracted from the three steganography algorithms. The dataset consists of 10,000 images in PGM format, which is converted to JPG format with a size of 256256 pixels. The data set is classified into clear and stego images, which are the same image samples used in each category, with the three separate data sets for stego images created using three algorithms (WOW, S-UNIWARD, and HUGO). The results show a slight decrease in detection accuracy, but the fine-tuning of the improved deep-CNN architecture performs better than other methods.

Keywords—steganalysis, deep learning, steganography algorithms, binary classification, algorithm detection

Cite: Awab Qasim Karamanji, Asia S. Ahmed, and Ali F. Fadhil, "Comparative Deep Learning Models in Applications of Steganography Detection," Journal of Image and Graphics, Vol. 12, No. 3, pp. 312-319, 2024.

Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.