Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Plug-and-Hide: Provable and Adjustable Diffusion Generative Steganography

Jiahao Zhu, Zixuan Chen, Yi Zhou, Weiqi Luo, Xiaohua Xie

TL;DR

This paper identifies a fundamental trade-off between stego image quality, steganographic security, and extraction reliability within the DM-GIS framework and proposes PA-B2G, a theoretically invertible mapping between secret bitstreams and stego images.

Abstract

Diffusion model-based generative image steganography (DM-GIS) is an emerging paradigm that leverages the generative power of diffusion models to conceal secret messages without requiring pre-existing cover images. In this paper, we identify a fundamental trade-off between stego image quality, steganographic security, and extraction reliability within the DM-GIS framework. Drawing on this insight, we propose \textbf{PA-B2G}, a \textbf{P}rovable and \textbf{A}djustable \textbf{B}it-to-\textbf{G}aussian mapping. Theoretically, PA-B2G guarantees the reversible encoding of arbitrary-length bit sequences into pure Gaussian noise; practically, it enables fine-grained control over the balance between image fidelity, security, and extraction accuracy. By integrating PA-B2G with probability-flow ordinary differential equations (PF-ODEs), we establish a theoretically invertible mapping between secret bitstreams and stego images. PA-B2G is model-agnostic and can be seamlessly integrated into mainstream diffusion models without additional training or fine-tuning, making it also suitable for diffusion model watermarking. Extensive experiments validate our theoretical analysis of the inherent DM-GIS trade-offs and demonstrate that our method flexibly supports arbitrary payloads while achieving competitive image quality and security. Furthermore, our method exhibits strong resilience to lossy processing in watermarking applications, highlighting its practical utility.

Plug-and-Hide: Provable and Adjustable Diffusion Generative Steganography

TL;DR

This paper identifies a fundamental trade-off between stego image quality, steganographic security, and extraction reliability within the DM-GIS framework and proposes PA-B2G, a theoretically invertible mapping between secret bitstreams and stego images.

Abstract

Diffusion model-based generative image steganography (DM-GIS) is an emerging paradigm that leverages the generative power of diffusion models to conceal secret messages without requiring pre-existing cover images. In this paper, we identify a fundamental trade-off between stego image quality, steganographic security, and extraction reliability within the DM-GIS framework. Drawing on this insight, we propose \textbf{PA-B2G}, a \textbf{P}rovable and \textbf{A}djustable \textbf{B}it-to-\textbf{G}aussian mapping. Theoretically, PA-B2G guarantees the reversible encoding of arbitrary-length bit sequences into pure Gaussian noise; practically, it enables fine-grained control over the balance between image fidelity, security, and extraction accuracy. By integrating PA-B2G with probability-flow ordinary differential equations (PF-ODEs), we establish a theoretically invertible mapping between secret bitstreams and stego images. PA-B2G is model-agnostic and can be seamlessly integrated into mainstream diffusion models without additional training or fine-tuning, making it also suitable for diffusion model watermarking. Extensive experiments validate our theoretical analysis of the inherent DM-GIS trade-offs and demonstrate that our method flexibly supports arbitrary payloads while achieving competitive image quality and security. Furthermore, our method exhibits strong resilience to lossy processing in watermarking applications, highlighting its practical utility.
Paper Structure (15 sections, 2 theorems, 12 equations, 4 figures, 5 tables, 3 algorithms)

This paper contains 15 sections, 2 theorems, 12 equations, 4 figures, 5 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminary
  4. Our Method
  5. Theoretical Analysis
  6. Design of PA-B2G
  7. Adjustable PA-B2G
  8. Message Hiding and Extraction with PF-ODEs
  9. Experiments
  10. Experimental setup
  11. Normality Test
  12. Method Comparison
  13. Diffusion Model Watermarking
  14. Computational Complexity
  15. Conclusion

Key Result

Theorem 1

Given a message sequence $\mathbf{m}=m_1m_2\cdots m_k$, $\mathbf{g}_s=\mathcal{C}_g\circ\mathcal{C}_u(\mathbf{m})\sim\mathcal{N}(\mathbf{0},\mathbf{I})$.

Figures (4)

  • Figure 1: Examples of $l=2$ for modes I and II. Along the vertical axis in each figure, the red regions represent the neighborhoods of quantiles, and the blue regions represent the variance-correction intervals. Sampling within these regions is prohibited. The green regions are the intervals that we actually adjust. Our variance-preserving strategy in each iteration involves two key steps: (a) increasing $S^2_u$ by increasing $c_1$, and (b) decreasing $S^2_u$ by increasing $c_2$.
  • Figure 2: Averaged histograms of noise that rejecting $H_0$w.r.t. different payloads and noise sizes under $\Delta_g=0.02$.
  • Figure 3: Visualization results of the adjustment strategies adopted by kim2025diffusion and PA-B2G on CIFAR-10, FFHQ, and LSUN-Bedroom.
  • Figure 4: Visualization results of combining PA-B2G in mode I with Stable Diffusion on text-to-image and image-to-image tasks.

Theorems & Definitions (4)

  • Theorem 1
  • proof
  • Theorem 2
  • proof