DKiS: Decay weight invertible image steganography with private key

TL;DR

DKiS tackles the security challenge of publicly known image steganography by introducing a preset private key within an invertible neural network framework, augmented by a decay weight that progressively reduces secret information transferred to the host pipeline. The method uses DWT/IWT preprocessing, a DKiBlock built from invertible components, and a private-key encoding that relies on patch shuffling and random ±1 multipliers, with a loss L = $\lambda_c L_c + \lambda_s L_s$ guiding embedding and extraction. Experimental results on DIV2K and additional datasets, along with ablations and attack simulations, show strong container/secret extraction performance and robust defense against key-less attacks, while demonstrating practical applications in photo verification and media authentication. Overall, DKiS advances secure, private-key-enabled steganography with a mechanism to control information leakage and a clear path toward real-world deployment.

Abstract

Image steganography, defined as the practice of concealing information within another image, traditionally encounters security challenges when its methods become publicly known or are under attack. To address this, a novel private key-based image steganography technique has been introduced. This approach ensures the security of the hidden information, as access requires a corresponding private key, regardless of the public knowledge of the steganography method. Experimental evidence has been presented, demonstrating the effectiveness of our method and showcasing its real-world applicability. Furthermore, a critical challenge in the invertible image steganography process has been identified by us: the transfer of non-essential, or `garbage', information from the secret to the host pipeline. To tackle this issue, the decay weight has been introduced to control the information transfer, effectively filtering out irrelevant data and enhancing the performance of image steganography. The code for this technique is publicly accessible at https://github.com/yanghangAI/DKiS, and a practical demonstration can be found at http://yanghang.site/hidekey.

Figures (11)

• Figure 1: The workflow of image steganography with a private key.
• Figure 2: The overview of DKiS.
• Figure 3: The architecture of DKiBlock. Left: forward process. Right: inverse process.
• Figure 4: The outputs of secret pipeline.
• Figure 5: The SSIM, APD and PSNR values between each $x_s^i$ and the secret image. * denotes that $x_s^i$ has been aligned to the secret image through a linear transformation. This alignment ensures that $x_s^i$ matches the secret image in terms of mean and standard deviation. This figure shows that the similarity/distance between $x_s^i$ and secret image is getting less/more as $i$ increasing.