Research On Detection Of Hidden Information In Text

Steganography, also called information hiding, is such a kind of techniques that imperceptibly hides and transmits secret information data so as to avoid the suspicion of the third party, and hence, effectively ensures the security of the storage and transmission of secret information as compared with the traditional encryption techniques. However, when steganography is increasingly and widely used to make contribution to the information security, it, on the other hand, has been also utilized by attackers to cause immeasurable threat for national security, social stability, and economical development. The steganalysis techniques, including detection, extraction, recovery and destruction of hidden information etc., therefore, obtain the increasingly growing attention and development as the promising method against the data hiding schemes. Among all techniques in steganalysis the detection technique is most important, as it is the premise and basis of the extraction, recovery and destruction techniques.When compared with images, text files have been more frequently used in various applications. So far, the text steganography techniques have been sufficiently studied and become more and more practical, while by contrast, only a little effort has been dedicated to the detection of hidden information in text. Furthermore, the existing methods of detecting hidden information in text have some significant drawbacks needed to be addressed.This paper takes the study on the universal text detection method, which can synchronously detect various text steganographic methods, as the research focus, then systematically investigates the main problems in the current text detection technique for text steganography. Moreover designs a text steganalysis system combining with the analysis of recovery and extraction of hidden information in text. The main results of these researches are as follows:Firstly, facing the problem that it is difficult to design superior detection algorithms for the linguistic steganographic algorithms, we have studied the researches on the two kinds of linguistic steganographic methods. For the Mimic-model-based steganographic method, a detection method is proposed based on source redundancy. This method processes the text and the words in the text file as the m-order Markov source and the source symbols, respectively, and then computes the redundancy of the source. Through analyzing the relationship between the redundancy and the size of the text, the existence of hidden information can be determined. For the synonym substitution based steganographic method, we have presented a detection algorithm by analyzing the feature of synonym pairings'value in text.Secondly, aiming at detecting the existence of hidden information embedded by the font format and invisible character based steganographic algorithms, we have proposed a text-noise-based detection method with good detection accuracy. The proposed method divides the text into several different information planes such as font-format plane, form-similar-character plane, invisible character plane, and so on, and next calculates the noise's value of the attributes of these planes to determine whether the detected text contains hidden information or not.Thirdly, in order to solve difficulty of effectively detecting small amounts of hidden information inserted in a clustered manner into large cover texts, a novel text detection approach, which is based on Haar wavelet packet decomposition, is introduced in this study. The approach makes full use of the wavelet packet decomposition to analyze the feature signal extracted from the test texts and the characteristics of the special components at different decomposed levels to determine the existence of hidden information. Both theoretical analysis and experimental results prove that this method can effectively eliminate the effect on the detection precision brought about by the dispersive random noise, detect and locate small amounts of hidden information embedded in a clustered manner into large cover texts.In the light of the aforementioned researches, for hidden information in text, we propose a universal detection algorithm based on Haar wavelet packet decomposition, via analyzing the features of normal natural texts and stego-texts generated by the various stegonagraphic algorithms, which are based on invisible character, form-similar-character substitution, additional code and Mimic model. Both theoretical analysis and experimental results prove that this algorithm has the better universality, as it succeeds in detecting the existence of hidden information embedded by the aforementioned stegonagraphic algorithms. In particular, it has the lower false negative rate and false positive rate for the stegonagraphic algorithms based on invisible character, form-similar-character substitution, and additional code.Finally, with the researches on the above difficulty and combination with the analysis of recovery and extraction of hidden information in text, a text steganalysis system has been implemented. And a recovery method is put forward for the dictionary and syntax database used by the text stegonagraphic tools based on Mimic model. Further, we also make studies on the extraction and recovery methods for the hidden information embedded by the tools such as Wbstego, Crypto123. Now, the text steganalysis system has been employed into practical application. It has the advantage of taking lower false negative rate and false positive rate for comprehensively detecting the hidden information embedded by the almost all the existing text stegonagraphic methods. Even it can recover some kind of hidden information in the text.Theoretical analysis and experimental results have demonstrated that the detection methods proposed in this paper have superior detection performance, and they greatly improve the development of the text steganalysis.