We propose a novel GAN-based steganographic approach, ARES, in which the Diversified Inverse-Adversarial Training (DIAT) strategy and the Steganalytic Feature Attention (SteFA) structure are designed to train a robust steganalytic discriminator. Specifically, the DIAT strategy provides the steganalytic discriminator with an expanded feature space by generating diversified adversarial stego-samples; the SteFA structure enables the steganalytic discriminator to capture more various steganalytic features by employing the channel-attention mechanism on higher-order statistics. Consequently, the steganalytic discriminator can build a more precise decision boundary to make it more robust, which facilitates learning a superior steganographic cost function.
- Train ARES
python main.py
If you use ARES in your research or wish to refer to the results published here, please cite our paper with the following BibTeX entry.
@article{cui2024ares,
title={ARES: On Adversarial Robustness Enhancement for Image Steganographic Cost Learning},
author={Cui, Qi and Zhou, Zhili and Meng, Ruohan and Wang, Shaowei and Yan, Hongyang and Wu, QM Jonathan},
journal={IEEE Transactions on Multimedia},
volume={26},
pages={6542--6553},
year={2024},
publisher={IEEE}
}