.png){kind=icon}
.png){kind=icon}
Return
Robust Image Steganography in Real Social Networks Using Stable Diffusion
.png){kind=icon}
PRE
AI
DOI：10.1109/JIOT.2026.3677876.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Generative steganography (GS), as an emerging information-hiding technique, directly synthesizes stego images using generative models. This effectively avoids the statistical anomalies introduced by modification-based steganography, which alters cover images. However, existing GS methods often struggle to achieve zero-error extraction of the secret message when faced with image processing operations on real social networks. To address the issue, this article proposes a robust generative image steganography scheme based on the stable diffusion model and a spread spectrum (SS) mapping module. Our scheme utilizes a pretrained stable diffusion model. Specifically, the secret message is modulated via orthogonal codes and a Gaussian perturbation and then mapped into a latent space noise that conforms to a standard Gaussian distribution. Combined with text guidance, this process generates stego images with high visual quality. At the receiver side, even if the image undergoes lossy transmission in real social networks, the secret message can still be achieved via inverse mapping, leading to zero-error extraction. Experimental results demonstrate that the proposed scheme can achieve zero-error extraction on Xiaohongshu, Weibo, X, and QQ. Moreover, in terms of undetectability, the detection error rate <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$P_{E}$ </tex-math></inline-formula> approaches 0.5, which is equivalent to random guessing.
Keywords:
Diffusion model
generative image steganography
robust steganography
social networks
steganography
Journal
IF:
8.9
Papers:
1.4W
Citations:
7.8W