崔建明, 余茜, 刘铭.基于WGAN的生成式信息隐写方法研究[J].河南理工大学学报（自然科学版）,2023,42(3):146-152.

CUI J M, YU X, LIU M.An innovative generative information steganography method based on Wasserstein GAN[J].Journal of Henan Polytechnic University(Natural Science) ,2023,42(3):146-152.

基于WGAN的生成式信息隐写方法研究

崔建明1, 余茜1, 刘铭2

1.长安大学 信息工程学院，陕西 西安  710064;2.国家计算机网络应急技术处理协调中心，北京  100029

摘要:针对图像载体隐写方法容量较低的问题，提出一种基于Wasserstein距离的生成式对抗网络（Generative Adversarial Network，GAN）模型无载体图像生成式信息隐写方法。首先提出一种支持用户自定义的“特定信息-噪声”映射关系矩阵，在此基础上将特定信息二次编码为噪声，然后将噪声输入到已训练的WGAN模型中，最终输出包含特定信息的可视化图像。实验结果表明，提出的信息隐写方法可以有效提高信息隐写算法的抗检测攻击效能，提升生成式信息隐写容量，减少隐写图像数量。此外，该方法依托较为简单的网络结构即可实现，不过分依赖系统性能和资源，具有较好的工程实用性。

关键词:信息隐藏;生成式信息隐写;WGAN;无载体隐写算法;隐写容量

doi:10.16186/j.cnki.1673-9787.2021090051

基金项目:国家自然科学基金-新疆联合基金重点资助项目（U2003206）；国家自然科学基金资助项目（62106060）

收稿日期:2021/09/15

修回日期:2021/11/22

出版日期:2023/05/25

An innovative generative information steganography method based on Wasserstein GAN

CUI Jianming1, YU Xi1, LIU Ming2

1.School of Information Engineering，Chang’an University，Xi’an  710064，Shaanxi，China;2.National Computer Network Emergency Response Technical Team/Coordination Center of China，Beijing  100029，China

Abstract:Aiming at the low capacity of steganography method of image carrier， a generative carrier-free information steganography method based on Wasserstein distance was proposed in this paper，which core model was Generative Adversarial Network（GAN）.Firstly，a specific information-noise mapping matrix that supporting customization was proposed.On this basis，specific information was re-encoded as noise fragments and inputted into the pre-trained WGAN.Finally，the network outputted a visualization image that contained specific information.The experimental results showed that the proposed steganography method could effectively improve the anti-detection efficiency and reduce the number of steganographic images while increasing the capacity of the generative information steganography.In addition，the proposed method could be implemented based on a relatively simple network structure and did not depend too much on system performance and resources，and had preferable practicability.

Key words:information hiding;generative information steganography;Wasserstein GAN;carrier-free steganographic algorithm;steganographic capacity

基于WGAN的生成式信息隐写方法研究_崔建明.pdf