StegNet: Mega Image Steganography Capacity with Deep Convolutional Network

Abstract

Traditional image steganography often leans interests towards safely embedding hidden information into cover images with payload capacity almost neglected. This paper combines recent deep convolutional neural network methods with image-into-image steganography. It successfully hides the same size images with a decoding rate of 98.2% or bpp (bits per pixel) of 23.57 by changing only 0.76% of the cover image on average. Our method directly learns end-to-end mappings between the cover image and the embedded image and between the hidden image and the decoded image. We further show that our embedded image, while with mega payload capacity, is still robust to statistical analysis.

A Survey on Deep Convolutional Neural Networks for Image Steganography and Steganalysis

“…Volkhonskiys [60] & Shis [34] also concentrated on generating secure cover images for traditional steganography algorithms. In 2018 Steg et al [32] proposed deep learning framework for image steganography to embed secret information without any involvement of traditional steganographic frameworks. Balujaet al [33] & Steg et al [32] both doing the same job.…”

“…In 2018 Steg et al [32] proposed deep learning framework for image steganography to embed secret information without any involvement of traditional steganographic frameworks. Balujaet al [33] & Steg et al [32] both doing the same job. Although, the concealed image is a bit detectable on residual images of the generated embedded images.…”

“…While embedding proposed method exploits the gradient feature map to determine the flipping direction of the pixels, flipping is equivalent to 1. After SaiMa et al [68] another successful approach by Wu et al [32], to apply neural networks for image steganography to embed secret information in images aside any influence or involvement of traditional steganography algorithms. The proposed network boost image steganography embedding rate to an average of 23.57bpp (bit per pixel) by modify only around of 0.76% of input cover image.…”

A Comprehensive Tutorial and Survey of Applications of Deep Learning for Cyber Security

“…This is avoided by introducing the GANs and detailed experimental analysis shown for Texture synthesizing. CNN based image steganography is studied for cover image by [105]. With the aim to enhance the robustness of the CNN model in image steganography [106], GANs are employed and additionally to improve the invisibility by hiding the secret image only in the Y channel of the cover image.…”

SteganoCNN: Image Steganography with Generalization Ability Based on Convolutional Neural Network

“…Given the known stego-image and carrier image, our human visual system can easily distinguish the differences between the two images. The solution proposed by Ping Wu et al [ 9 ] solves the distortion of the stego-image, thereby ensuring the visual integrity of the stego-image. Zhang et al [ 10 ] decomposed the carrier image (color image) into three channels corresponding to the Y, U, and V channels.…”