Research On Steganalysis For Empty Word Modification Based Steganography In Text

In the Internet age, the research on text steganography has become an important means of covert communication. However, the abuse of this technology may pose a serious threat to national security and social stability. Consequently, it is an urgent task to research the detection of hidden information.Text files have been frequently used in various applications. So far, the text steganography techniques have been sufficiently studied and become more and more practical. Natural language steganography represents the trends of text steganography. Semantic steganography is an important component of natural language steganography, which possesses nice anti-detection characteristic. The existing steganalysis methods aiming to semantic steganography are mostly only effective for synonym substitution based steganographic method.This paper takes the study on the steganalysis method for the function word modification--based steganography in text, which is belong to semantic steganography. Comparing to synonym substitution based steganographic method, it possesses nice imperceptibility characteristic. This paper first proposes a method suitable for single function word modification--based steganography, and then it researches a general method that can also deal with multiple function word modification--based steganography. The main results of these researches are as follows:By investigating the characteristics of function words and analyzing the changes in the number of different function words categories after being embedded, this dissertation proposes a steganalysis method for the single function word modification--based steganography in text. The method utilizes an algorithm based on the nearest neighbor for statistical detection and combines it with syntax analysis. The experimental results show that the false positive rate and the false negative rate of this steganalysis method both are low.On the basis of research on steganalysis method for the single function word modification--based steganography, this dissertation then proposes a steganalysis method, which is effective both for single and multiple function words modification--based steganography. This method first establish the channel model of function word modification--based steganography, so as to analyze the statistical dependency of input signal and output signal, then analyzing hopping value of function words in text, at last train samples to get the classification interface using perceptron algorithm. Through the interface, the existence of hidden information can be determined.