PSyDUCK: Training-Free Steganography for Latent Diffusion

2025-01-31

AI Summary

To address privacy-preserving communication needs of vulnerable field operatives—such as journalists and humanitarian responders—in repressive environments, this paper proposes a training-free, cross-model-compatible steganographic framework operating in the latent space of diffusion models (e.g., Stable Diffusion, SVD). Our method leverages geometric analysis of latent variables and adaptive perturbation-based encoding, embedding secret messages via forward/reverse process invariance—without model weight modification or fine-tuning. It supports both image and video generation tasks, overcoming traditional steganography’s constraints of model lock-in and training dependency. Experiments demonstrate high payload capacity, imperceptibility to human observers, decoding accuracy ≥99.2%, and strong robustness against common distortions and model variations.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Web Mining and Content Analysis: Large pretrained models with web data
• Security and Privacy: Data transparency and provenance

Abstract

Recent advances in AI-generated steganography highlight its potential for safeguarding the privacy of vulnerable democratic actors, including aid workers, journalists, and whistleblowers operating in oppressive regimes. In this work, we address current limitations and establish the foundations for large-throughput generative steganography. We introduce a novel approach that enables secure and efficient steganography within latent diffusion models. We show empirically that our methods perform well across a variety of open-source latent diffusion models, particularly in generative image and video tasks.

Problem

• Information Hiding
• Efficiency
• Security

Innovation

• PSyDUCK
• Information Hiding
• Stealth Diffusion