Abstract: As overall statistical characteristics of the cover text would be changed in the process of synonym substitutions, this paper proposes a text steganographic algorithm using the prior substitutions with low distortion. Taking the suitability of synonyms in the context into account, the proposed algorithm constructs a distortion function to measure the degree of the statistical characteristics changed by synonym substitutions. It not only selects two words with the highest suitability from each synonymous word set to form a substitution set, but also from the global perspective, gives priority to the substitution sets with low distortions to be embedded secret information after sorting the distortions caused by all the substitution sets in the cover text. Therefore, the changed degree of the statistical characteristics is reduced. The experimental results show that the algorithm can effectively resist the attacks made by detection algorithms using statistical features based on the synonyms pairing and relative frequency. The proposed algorithm has a good resistance to detection, and can improve the security of the secret information.

Key words: text steganography, synonym substitution, distortion function, steganalysis

摘要： 针对同义词替换操作造成原始文本整体统计特性的破坏，提出了一种基于低失真替换优先的文本隐写算法。该算法以同义词在文本中的合适度为基础，构造失真函数衡量同义词替换后文本统计特性的改变程度；不仅为每个同义的词集合选取合适度最高的两个词组成替换组合，而且从全局的角度出发，对原始文本中的所有同义词替换组合所引起的失真度进行排序，优先选取失真度小的同义词替换组合来嵌入信息，从而降低了同义词序列统计特性的改变程度。实验结果表明，该算法可以很好地抵抗基于同义词结对和基于同义词相对词频统计特征的隐藏信息检测算法的攻击，具有较高的抗检测能力，提高了秘密信息的安全性。

关键词: 文本隐写, 同义词替换, 失真函数, 隐写分析