Generative Image Steganography Algorithm Based on Diffusion Model and Style Transfer

doi:10.1109/tdsc.2026.3737790

Generative Image Steganography Algorithm Based on Diffusion Model and Style Transfer

2026-09-25

0

PRE

AI

DOI：10.1109/tdsc.2026.3737790

Original

Original request for help

Save

Abstract

En 中文

Generative image steganography based on style transfer is a technique that generates stego images in the process of image style transfer. However, most of the existing methods that use Generative Adversarial Networks (GANs) suffer from serious problems of insufficient embedding capacity and extraction accuracy due to the limited generation ability of GANs itself. To overcome this issue, this paper proposes a novel generative image steganography based on diffusion model and style transfer. Our algorithm introduces three key innovations. Firstly, a diffusion based steganography framework based on style transfer is introduced to generate the stego transfer image with high visual quality. Secondly, a binary interval mapping mechanism based on position switching is designed to map the secret information into high-dimensional feature, thus improving the embedding capacity and extraction accuracy. Thirdly, a steganographic path that treats content feature as carriers in style transfer is proposed to further enhance the stability of secret information extraction. This triple strategy pioneers to embed a large amount of secret information during the style transfer process with the diffusion model, enabling the generation of high-quality stego images. Experimental results demonstrate that compared with existing generative steganography methods based on style transfer, our method achieves superior performance in terms of visual quality, extraction accuracy and embedding capacity. This algorithm shows good potential in covert communication.

Keywords:

Generative image steganography

Diffusion model

Style transfer

Binary interval mapping

Position switching

Generative Image Steganography Algorithm Based on Diffusion Model and Style Transfer

Abstract

Journal

Organization

Cited Papers