CCCIH: Content-consistency Coverless Information Hiding Method Based on Generative Models

Document Type

Article

Publication Date

12-1-2021

Abstract

In order to improve the embedding capacity and security of coverless information hiding methods, a content-consistency coverless information hiding method based on generative models is proposed. In this letter, we point out the generative models, the most appealing framework for image generation is suitable for coverless information hiding. Starting from the initial concept (i.e. No Cover) of coverless information hiding, our model is composed of two generative models: F and G, which are utilized to generate cover image and reconstruct secret image, respectively. The attempt to utilize G to reconstruct secret image usually leads to color distortion and content loss. We realize that the problem is due to the lack of content information of secret image in the middle image, which is generated by F. An extraction module is added during the generation of cover image, which is called “content-consistency”. The experimental results clearly verify the capacity of our model, which can further improve the quality of reconstructed secret image. Moreover, compared with other coverless information hiding methods, the embedding rate of our model is much better than other methods.

Identifier

85110447885 (Scopus)

Publication Title

Neural Processing Letters

External Full Text Location

https://doi.org/10.1007/s11063-021-10582-y

e-ISSN

1573773X

ISSN

13704621

First Page

4037

Last Page

4046

Issue

6

Volume

53

Grant

MMJJ20170203

Fund Ref

National Natural Science Foundation of China

This document is currently not available here.

Share

COinS