一种基于GAN和纠错编码技术的无嵌入隐写方法
投稿时间:2020-08-14  修订日期:2020-10-19  点此下载全文
引用本文:
摘要点击次数: 40
全文下载次数: 0
作者单位E-mail
梁天一 华东理工大学 理学院 tianyilt@qq.com 
梁谦旺 华东理工大学 信息科学与工程学院  
施秦 华东理工大学 信息科学与工程学院  
魏苏航 华东理工大学 信息科学与工程学院  
蒋翠玲 华东理工大学 信息科学与工程学院 cuilingjiang@ecust.edu.cn 
基金项目:国家自然科学基金资助项目(No. 61872143)。华东理工大学国家级大学生创新创业计划资助项目(No. S19089);
中文摘要:传统的隐写方法是将秘密信息嵌入到载体中,而利用生成式对抗网络(GAN)的无载体信息隐藏技术逐渐成为目前最有潜力的研究方向。本文在一种基于GAN的无嵌入信息隐藏方法的基础上,进行了研究与改进。信息发送方首先利用噪声驱动生成器直接生成含密图像,信息接收方通过秘密信息提取器提取出有效的秘密消息,改进提取器的训练任务,并引用冗余纠错编码技术以提高信息提取的准确率。实验结果表明,相比同类方法,本文方法在保持高隐写容量的情况下,具有更高的信息提取准确率,同时收敛速度大大提高。引入纠错码技术大幅提高了秘密信息提取的准确率,具有一定的实用性。
中文关键词:信息隐藏  无嵌入隐写  生成式对抗网络  纠错编码
 
A Steganography method without embedding based on GAN and error-correcting code technology
Abstract:Traditional image steganography algorithms embed a secret message into a cover image. However, steganography without em-bedding, which uses Generative Adversarial Networks (GANs), has become the most potential direction in information hiding at present. We studied and improved a steganography method without embedding based on GAN. First, the sender uses a generator as the noise driver to generate the cover image. Next, the receiver extracts the secret information using the extractor, and then com-pletes secret communication. In the process of extracting information, improvement of the extractor training task and redundant correcting code technique is utilized to improve information extraction accuracy. Experimental results show that, compared with similar algorithms, the proposed method ensures higher information extraction accuracy while maintaining high steganography capacity. Meanwhile, the convergence speed of training is also greatly improved. The introduction of error-correcting code technology greatly improves the accuracy of secret information extraction and has certain practicability.
keywords:information hiding  steganography without embedding  generative adversarial network  error correcting codes
查看全文   查看/发表评论   下载pdf阅读器