Abstract—Steganography, or hiding information within digital media, is one of the most important challenges in digital security in terms of detecting hidden content for both various embedding processes and under different payload sizes. This study proposes an enhanced deep learning methodology that combines the Visual Geometry Group 19-layer Convolutional Neural Network (VGG19) convolutional neural network with particle swarm optimization to optimize key hyperparameters, improving its ability to detect steganographic content more effectively. Our proposed approach was tested using the Break Our Steganographic System (BOSSBase) 1.01 dataset and a combined dataset with Break Our Watermarking System 2 (BOWS2), focusing on stego-images generated by the Spatial UNIversal WAvelet Relative Distortion (S-UNIWARD) and Wavelet Obtained Weights (WOW) algorithms. The results clearly indicate that our proposed methodology outperforms state-of-the-art models such as Xu-Net, Ye-Net, Yedroudj-Net, and VGG16Stego, achieving accuracy of 0.8816 and 0.8900 for payloads of 0.2bpp (bits per pixel) and 0.4bpp, respectively. These findings show the significance of our approach, highlighting its potential to become a leading solution for steganography detection in digital security applications.
 
Keywords—steganalysis, particle swarm optimization, Visual Geometry Group 19 (VGG19), spatial domain, image security, data hiding


Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).