Advances in Steganography and Watermarking

The field of steganography and watermarking is moving towards developing more sophisticated and robust methods for hiding and detecting secret information in images and videos. Recent research has focused on improving the payload capacity of steganographic schemes, with some studies achieving significant increases in capacity while maintaining image quality. Additionally, there is a growing interest in developing methods for proving authorship and protecting copyright in generated content, particularly in the context of private diffusion models. Noteworthy papers in this area include: We Can Hide More Bits, which establishes upper bounds on the message-carrying capacity of images and demonstrates that current methods have not yet saturated watermarking capacity. NoisePrints, which proposes a lightweight watermarking scheme for private diffusion models that utilizes the random seed used to initialize the diffusion process as a proof of authorship. Foveation Improves Payload Capacity in Steganography, which improves existing capacity limits in steganography using foveated rendering and efficient latent representations.

Sources

Targeted Pooled Latent-Space Steganalysis Applied to Generative Steganography, with a Fix

We Can Hide More Bits: The Unused Watermarking Capacity in Theory and in Practice

Reference-Specific Unlearning Metrics Can Hide the Truth: A Reality Check

Foveation Improves Payload Capacity in Steganography

NoisePrints: Distortion-Free Watermarks for Authorship in Private Diffusion Models

LOTA: Bit-Planes Guided AI-Generated Image Detection

Nonparametric Data Attribution for Diffusion Models

Rate-Adaptive Spatially Coupled MacKay-Neal Codes with Thresholds Close to Capacity

Rate-Adaptive Protograph-Based MacKay-Neal Codes

Built with on top of