Abstract: In order to enhance the statistical security of a data hiding system, the factors that influenced the security of a steganography were analyzed. Three ways to reduce the statistical distortion of the stego-image were found: Increasing embedding efficiency, controlling modifying amplitude and choosing embedding position adaptively. According to the three ways, a new adaptive steganography based on double-layered embedding method was proposed. The new scheme chose pixels under strong noise background as message carrier and embedded secrets in their least significant and second least significant bit-planes. The experimental results on uncompressed images show that the proposed steganography outperforms the prior algorithm for resisting blind steganalysis.

Key words: adaptive steganography, local correlation, embedding model, resisting blind steganalysis, double-layered embedding

摘要： 为了提高隐写系统抗统计检测的能力,对影响隐写算法统计安全性的因素进行了分析,得出减小载密图像统计失真的3条途径。以此为基础,提出一种基于双层湿纸嵌入的图像自适应隐写算法。该算法根据邻域相关性,选取局部环境噪声强的像素作为嵌入对象,利用“湿纸码(WPC)”在其最低有效位及次低有效位嵌入秘密信息。对未压缩图像库的实验表明,该方法比现有算法具有更好的抗统计检测性能。

关键词: 自适应隐写, 邻域相关性, 隐写模型, 抗盲检测, 双层嵌入