Image Steganography Using Deep Learning Techniques

Anthony Rene Guzman, Purdue University

Abstract

Digital image steganography is the process of embedding information within a cover image in a secure, imperceptible, and recoverable way. The three main methods of digital image steganography are spatial, transform, and neural network methods. Spatial methods modify the pixel values of an image to embed information, while transform methods embed hidden information within the frequency of the image. Neural network-based methods use neural networks to perform the hiding process, which is the focus of the proposed methodology. This research explores the use of deep convolutional neural networks (CNNs) in digital image steganography. This work extends an existing implementation that used a two-dimensional CNN to perform the preparation, hiding, and extraction phases of the steganography process. The methodology proposed in this research, however, introduced changes into the structure of the CNN and used a gain function based on several image similarity metrics to maximize the imperceptibility between a cover and steganographic image. The performance of the proposed method was measured using some frequently utilized image metrics such as structured similarity index measurement (SSIM), mean square error (MSE), and peak signal to noise ratio (PSNR). The results showed that the steganographic images produced by the proposed methodology are imperceptible to the human eye, while still providing good recoverability. Comparing the results of the proposed methodology to the results of the original methodology revealed that our proposed network greatly improved over the base methodology in terms of SSIM and compares well to existing steganography methods.

Degree

M.Sc.

Advisors

Inukollu, Purdue University.

Subject Area

Artificial intelligence|Electrical engineering|Mathematics

COinS