Hang Yang

Hang Yang

he/him

PhD candidate in Mathematics at UMass Amherst. Researching machine learning, robotics, and deep learning — including mobile motion capture with robot dogs, invertible neural networks for image steganography, and vision transformers for medical AI.

DKiS: Decay weight invertible image steganography with private key

Published in Neural Networks, 2025

We propose DKiS, an invertible image steganography framework that addresses two key limitations of existing methods:

  1. Decay Weight Mechanism — regulates the transfer of non-essential “garbage” information from the secret to the host pipeline, improving steganographic quality and robustness.
  2. Private Key Integration — incorporates a preset private key into high-capacity image steganography for the first time. Access to concealed data requires the corresponding private key, significantly strengthening security against adversarial extraction.

DKiS provides a practical solution for secure, high-capacity image steganography applicable in real-world information hiding scenarios.

Code available at https://github.com/yanghangAI/DKiS.

Recommended citation: Yang, H., et al. (2025). "DKiS: Decay weight invertible image steganography with private key." Neural Networks, 185, 107148.
Download Paper