Hiding Images into Images with Real-world Robustness

TL;DR

This paper presents a generative deep network approach for embedding images into other images with high robustness against real-world attacks like JPEG compression and noise, enabling effective copyright protection.

Contribution

It introduces a novel deep network architecture with an attack layer and decoupling network, improving robustness and allowing the hiding of three secret images, surpassing previous methods.

Findings

Outperforms existing methods against digital attacks

Achieves high-quality image recovery after attacks

First to robustly hide three secret images

Abstract

The existing image embedding networks are basically vulnerable to malicious attacks such as JPEG compression and noise adding, not applicable for real-world copyright protection tasks. To solve this problem, we introduce a generative deep network based method for hiding images into images while assuring high-quality extraction from the destructive synthesized images. An embedding network is sequentially concatenated with an attack layer, a decoupling network and an image extraction network. The addition of decoupling network learns to extract the embedded watermark from the attacked image. We also pinpoint the weaknesses of the adversarial training for robustness in previous works and build our improved real-world attack simulator. Experimental results demonstrate the superiority of the proposed method against typical digital attacks by a large margin, as well as the performance boost of the recovered images with the aid of progressive recovery strategy. Besides, we are the first to robustly hide three secret images.