V-Net architecture based advanced visually meaningful image encryption technique

Abstract

The visually meaningful image encryption (VMIE) technique is used to offer additional security to the noise-like encrypted image by implanting the secret image in the cover image. Generally, VMIE techniques follow two steps: pre-encryption and embedding. In this paper, an advanced V-Net architecture based VMIE technique is presented. V-Net architecture is based on a Convolution Neural Network (CNN) and offered medical image segmentation, that is utilized in the image into image steganography in the proposed work. In the proposed VMIE technique, the secret image is encrypted by utilizing the 5D chaotic map to provide high security, and the encrypted secret image is implanted into the cover image using a V-Net based embedding method. CNN-based image embedding methods offer high payload capacity and imperceptibility. In these methods, the encoder hides the secret image into the cover and the decoder reconstruct the secret image. Furthermore, a CNN-based efficient decoder is designed in the proposed method. Moreover, to validate the proposed method, standard evaluation parameters are analyzed.