知识图谱引导的内容可控的文本生成式隐写方法研究

文本生成式隐写方法能够生成接近人类交流所使用的自然语言文本，具有较高的隐蔽性，是信息隐藏领域的热门研究方向。当前文本生成式隐写方法主要基于文本生成模型和条件概率编码实现。秘密信息的嵌入将不可避免地破坏原始文本载体的统计分布特征，从而降低生成文本质量，并且生成的隐写文本不具有内容上的相关性和连贯性，文本内容不可控，容易被第三方检测识别，将极大影响隐写算法的隐蔽性。针对上述问题，本项目尝试利用知识图谱引导隐写文本的语义表达，在文本生成过程中融入结构化知识，建立一种内容可控的文本生成式隐写新框架。并在此框架基础上，研究面向知识图谱的语义计算和语义匹配、唯一性路径编码、基于图嵌入模型的文本生成方法等关键技术，实现包含特定内容的秘密信息隐藏与提取。本项目研究以生成内容可控的高质量隐写文本为目标，旨在进一步提升网络大数据环境下隐写方法的隐蔽性，为隐写技术应用提供技术支撑，具有较高的研究价值和现实意义。

Generation-based linguistic steganography which can generate natural language has high concealment. It has become a popular research direction in the field of information hiding. The current generation-based linguistic steganography framework is mainly composed of two parts: text generation model and conditional probability coding. Thanks to the application of machine learning in the field of natural language processing, the quality of the generated text has been significantly improved. However, these methods which based on conditional probabilistic coding have the inherent defect that they cannot simultaneously increase the hiding capacity and improve the concealment. Furthermore, the generated text is not relevant and coherent, which makes it easy to be detected by others. To solve these problems, we use the knowledge graph to guide the semantic expression of the steganographic text and a new framework of generation-based linguistic steganography method is proposed. Then, the key technologies including semantic representation and computing, unique path coding, and graph encoding are studied to implement algorithms of information hiding and extraction. This project aims to improve the concealment of steganography methods and provide technical support for the application of information hiding technology, which has high research value and practical significance.