A Dual Secured Medical Image Steganography Model to Enhance Network Security based on Deep Learning Techniques

A growing number of researchers have taken an interest in steganography due to the value it offers to information security over the past decade. A process named medical image steganography conceals confidential medical data within an image. It is common practice to securely embed secrets in image steganography methods so that the payload capacity is almost forgotten and the human visual system quality is not good enough. By converting the cover into frequency domain with DWT, the high frequency components are optimally selected with pixels. For transmitting secure data, Lempel-Ziv Welch (LZW) and Huffman techniques are used first. In the next step, the data is encrypted with RC4 encryption. The encrypted data into the cover image is accomplished through the hidden network (H-net). In real life applications, deep learning-based image steganography is relatively rare. This research proposes a novel Convolution Neural Network based on H-net and R-net model that can successfully recover secret data, while solving the challenge of secret images embedded in a carrier image. The network is trained throughout its entirety, from start to end. Then, the secret picture is embedded into the carrier by the encoding network, and the distinct secret images are reconstructed by the decoding network. Quality of stego image is further improved by using HFNN.