基于自动选择编码及动态选词策略的文本隐写方法  

Text steganography method based on automatic selection coding and dynamic word selection strategy

作　　者：李晖[1] 金家立 金纾羽 马卫娇 LI Hui;JIN Jiali;JIN Shuyu;MA Weijiao(School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China;Beijing YuanliWeilai Science and Technology Co.,Ltd,Beijing 100102,China;School of Computing,Neusoft Institute Guangdong,Foshan 528225,China)

机构地区：[1]沈阳工业大学信息科学与工程学院,辽宁沈阳110870 [2]北京猿力未来科技有限公司,北京100102 [3]广东东软学院计算机学院,广东佛山528225

出　　处：《通信学报》2022年第9期240-253,共14页Journal on Communications

摘　　要：针对文本编码方式不灵活以及候选词增加导致生成的隐写文本质量较低的问题,提出了一种基于自动选择编码及动态选词策略的文本隐写方法。所提方法基于Transformer的神经机器翻译模型生成隐写译文。在生成隐写译文的过程中,采用定长编码和哈夫曼编码建立候选词与码字之间的映射关系,通过计算隐写词元与正常词元的概率差异百分比,实现基于概率差异阈值的动态选词。最后,比较生成的2种隐写译文Sacrebleu的大小,实现编码方式的自动选择。实验结果表明,所提方法能够生成流畅度高、可读性强的隐写译文。当隐藏容量为11.19%时,隐写译文的Sacrebleu达到10.53。A text steganography method based on automatic selection coding and dynamic word selection strategy was proposed for the inflexible text coding method and candidate word increasing number leading to the low quality of generated steganographic text. Steganographic translations was generated based on Transformer’s neural machine translation model. In generating steganographic translations, fixed-length coding and Huffman coding were used to establish the mapping relationship between candidate words and codewords, and dynamic word selection based on the probability difference threshold was achieved by calculating the probability difference percentage between steganographic words and normal words. Finally, the size of the two generated steganographic translations Sacrebleu was compared to realize the automatic selection of coding mode. The experimental results show that the proposed method can generate steganographic translations with high fluency and readability. When the embedding rate is 11.19%, the Sacrebleu of the steganographic translation reaches 10.53.

关 键 词：信息隐藏 自然语言生成 文本隐写 机器翻译 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]