基于HUGO-WGAN图像信息隐写方案  

Image Information Steganography Scheme Based on HUGO-WGAN

作　　者：张菊 阚凌宇 王思纯 杨起豪 ZHANG Ju;KAN Lingyu;WANG Sichun;YANG Qihao(Huaqiao University,Xiamen Fujian 361021)

机构地区：[1]华侨大学,福建厦门361021

出　　处：《软件》2021年第4期81-83,95,共4页Software

基　　金：华侨大学2020年度大学生创新训练计划项目(S202010385025)。

摘　　要：信息隐藏是将信息以不可见的方式隐藏在一个宿主信号中,以实现机密通信和版权保护等目标。它主要用于两个特定的秘密渠道,尤其是在快速发展的社交网络中,以海量的图像和视频为媒介。这为隐藏信息提供了更多的机会和挑战,对国家安全和信息安全都十分重要。基于传统图像信息隐藏方法存在的难以抵抗基于特征学习的深度神经网络隐写分析技术的问题,本项目将生成式对抗网络引入到原始图像隐写方法中[1],利用GAN(Generative Adversarial Networks)模型中的博弈对抗策略,为图像信息隐藏生成更安全的载体信息,更高视觉质量的图像,增强其抵抗隐写分析的性能。提出了一种新的隐写方案,调用WGAN(Wasserstein GAN)模型框架,采用HUGO(Highly Undetectable ste GO)隐写算法实现更加安全的图像隐写操作[2]。Information hiding is to hide information in a host signal in an invisible way to achieve the purpose of covert communication and copyright protection.It is mainly used for secret channels between specific parties,especially in the fast-growing social network,with rich images and videos as carriers,which provides more opportunities and challenges for information hiding,and is important for national security and information security.Meaning.Based on the traditional image information hiding method that is difficult to resist the problem of feature learning-based deep neural network steganalysis technology,this project introduces the generative confrontation network into the original image steganography method,using the GAN(Generative Adversarial Networks)model The game confrontation strategy generates safer carrier information for image information hiding,higher visual quality images,and enhances its resistance to steganalysis.A new steganography scheme is proposed,which calls the WGAN(Wasserstein GAN)model framework and uses the HUGO(Highly Undetectable steGO)steganography algorithm to achieve more secure image steganographyoperations.

关 键 词：图像隐写 生成对抗网络(GAN) WGAN HUGO 安全性 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]