Deep learning driven image steganalysis approach with the impact of dilation rate using DDS_SE-net on diverse datasets

Abstract

The challenge of effective and precise steganalysis is crucial in the field of digital steganography. Steganalysis is a constantly evolving field of study that looks for hidden data in digital media. With the recent developments in communication and information technology, as well as information law compliance, image Steganalysis has drawn a lot of attention. The methods for steganography that are now available make it harder to identify steganographic material. This study presents a comprehensive investigation of the DDS_SE-Net architecture based on convolution neural networks employing various datasets in steganalysis using key performance measures, including accuracy, recall, precision, and F1-score. Additionally, this study looks at how rate of change of dilation in DDS_SE-Net contributes to the improved outcomes. In this work dilation rate of 3 gave comparatively better accuracy of 92.9% against WOW, 89.2 and 89.8% against S-UNIWARD and HILL, respectively. The results show that the deep learning framework selected and the data used in training have a major impact on how well the model performs steganalysis.