Natural Language Analysis And Steganography Steganography Amount Detection

Text steganography is an important way to covert information through the Internet. As one branch of text steganography, natural language steganography has been paid attention by more and more people. However, in recent years, as the natural language steganography matures gradually, criminals make use of this technology to disseminate the illegal information which endanger national security through the Internet. This phenomenon has become the risk of public security and national defense. Therefore, the natural language steganalysis is already brook no delay!Natural language steganalysis is used to detect whether a text contains secret information. The research of this technology not only can monitor the transmission of network information effectively, but also can promote the development of existing steganography technologies. At present, the study of natural language steganalysis is not much, and the depth of research is also limited. So, natural language steganalysis cannot meet the needs of practical applications and it is difficult to play the role of promoting the development of the existing technologies.From the view of statistics, this paper proposes a natural language steganalysis and steganography quantity model which not only can detect whether a text contains secret information, but also can estimate the steganography quantity and the upper bound of the steganography quantity. The main research contents in this paper are:1) Proposes the natural language steganalysis model. The model base on statistics, take text feature replacement rate distribution as the classification features to detect the text if it contains secret information. This paper verifies the feature replacement rate of normal texts is beta distribution and the feature replacement rate of embedded texts is normal distribution whose mean is0.5from theoretical aspect and experimental aspect. Experiments part discussed the influence of text length and feature threshold to model accuracy.2) Proposes natural language steganography quantity detection model. This paper proves that the feature replacement rate distribution of different text length, different feature threshold and different steganography quantity are mixed distribution combine with beta distribution and normal distribution, thus proposes natural language steganography quantity detection model.3) Combined the beta distribution of normal texts and the mixed distribution of embedded texts, this paper proposes a upper bound model of steganography quantity. This model utilize beta distribution and mixed distribution to fit the feature replacement rate distribution of normal texts and embedded texts respectively, then compute the security rate and plot the curve of steganography quantity upper bound.