Implementation and Comparative Analysis of CNN and Discrete Haar Wavelet Transform in Image Steganography

In the developing field of Steganography, there is a constant search for more secure and efficient methods of hiding secret images within digital images. This paper presents the implementation and comparative analysis of the traditional Discrete Wavelet Transform (DWT) steganography technique using Haar wavelets with the modern Convolutional Neural Network (CNN)-based method. Both the models are compared using metrics such as Structural Similarity Index (SSI), Mean Squared Error (MSE), and Peak Signal-to-Noise Ratio (PSNR) to determine the efficiency and robustness in effectively hiding the secret image. The results show an improvement in SSI by 8.42%, improvement in PSNR by 58.46%, and decrement in MSE by 85.05%, implying that CNN based steganography method significantly outstrips the traditional Discrete Haar wavelet based steganography technique. The CNN approach not only enhances the undetectability and security of the hidden information but also maintains higher accuracy of the steganographic images, thus marking an important advancement in the domain of digital steganography.