Generative feedback residual network for high-capacity image hiding

Abstract

 At present, there is always a potential threat in the process of information transmission. As a way to protect data security, image hiding has attracted extensive attention. Current image hiding algorithms have insufficient resistance to deep leaning based steganalysis algorithms and relatively low hiding capacity. This paper presents an image hiding algorithm based on a generative feedback residual network (GFR-Net), which hides multiple color secret images in a single color carrier image. First, several secret images and a carrier image were fed into the image hiding network, in which the secret images were embedded into the carrier image, resulting in an output of container image. A recovery network also based on GFR-Net was designed to reconstruct the secret images from the container. The extensive experiments for hiding normal and encrypted images show that the proposed image hiding model has a good performance in terms of payload and security.

KEYWORDS:

• Image hiding
• generative image hiding
• convolutional neural network
• feedback residual

Funding

This work was partly supported by the National Natural Science Foundation of China (Grant No. 62041106).

Disclosure statement

No potential conflict of interest was reported by the author(s).