Research And Implementation Of Natural Language Information Hiding Algorithm Based On Abstract Embedding Unit

With the rapid development of network and communication technology and digital products are widely used. Information Hiding (IH) technology has become an important field of modern information security. Natural language is the main way for people exchanging their information, but the existing information hiding research results of images, videos, et al cannot meet the needs of Natural Language based text Information Hiding (NLIH). So this paper researches on NLIH technology.By analyzing the current NLIH technologies, we find out that the existing algorithms are all insufficient of hiding capacity and the cover process combined closely with coding algorithms, which make the algorithm less versatile.Based on the above issues, we proposed an NLIH algorithm based on Abstract Embedding Unit (AEU), IH algorithm can be divided into two parts by AEU, one is cover process, and the other is coding algorithm. By using AEU as the embedding unit, the coding algorithm only needs to focus on the process of information embedding and extracting. It can not only enrich the hiding capacity, but also can improve the universal of coding algorithm.In cover process part, in order to get enough AEU, we design cover process algorithms of synonyms substitution and syntactic transformation. In coding algorithm part, we propose a coding algorithm of coloring the graph. The current synonyms based NLIH algorithm cannot avoid the word sense disambiguation error when extract the hiding message. We use the key to coloring the graph nodes which indicate the synonyms and sentence templates. The colors include blue, green and black. Embedding and extracting with the same key, so all the nodes' color will not change. When extracting, we just get the word color in the graph and then decode it; this can highly improves the accuracy of extraction.Finally, we design and implement a prototype system of NLIH. We implement embed and extract experiments by using English corpus as testing text. Experiments show that the proposed algorithm has better robustness and concealment, and the hiding capacity is improved.