Enhancing Covert Communication Systems through Deep Steganography

Abstract

Deep image steganography has emerged as a promising approach for concealing secret information with in digital images, leveraging the power of deep learning techniques to enhance security. The study explores the integration of deep neural networks to enhance the security and robustness of steganographic methods. Conventional image steganography approaches may have vulnerabilities because of their fixed algorithms, which makes them less flexible in handling different types of image material and more prone to being detected by advanced steganalysis methods. The proposed method aims to advance covert communication systems by utilizing deep learning to achieve imperceptibility and robustness against common structural analysis methods.

This study presents a generic Convolutional Neural Network (CNN) with encoder-decoder architecture for deep image steganography method. It facilitates the seamless concealment and extraction of information. In order to assess the efficacy of the proposed approach, peak signal-to-noise ratio and structural similarity index measurements were used. Experimental results based on an ImageNet dataset show that our approach outperforms the selected related methods in terms of security, robustness and visually imperceptibility. PSNR: 72.79 SSIM: 0.9753 The test results, thus gleaning that the approach we proposed is superior to previous methods.