Research On Analysis And Design Of Linguistic Steganography

Traditional cryptography turns a plain message into an incomprehensible ciphertext through the implementation of various transformations, providing a powerful means of the protection of secret information. However, cryptography has its inherent shortcomings. It exposes the important information such as the security and importance of the message itself, the sender and the receiver's information while applying transformations to disturb the message, which may attract the attention of the third-party and be attacked. Moreover, some governments have limited the use of cryptography. Information hiding technique can overcome these shortcomings and avoid the restrictions from the governments desirably. Information hiding technique conceals the very existence of secret information, which greatly enhances the security of storation and transmission of information.Text information hiding is a kind of information hiding that uses text as its carrier. It hides secret information in the text carrier using the redundancy in format, coding, structure, syntax and semantics. In the Internet, text information has played a very important role. As the text data is very large and text processing is more intuitive, the use of text information hiding is an attractive idea.Linguistic steganography is a subclass of text information hiding. It uses the redundancy of the natural language content in text to hide secret information, for the main purpose of secure communications. Compaered to other kinds of text information hidding, it has higher security and better robustness, and thus attracts more attention of researchers. Linguistic steganography is an organic integration of information hiding and natural language processing.Linguistic steganalysis refers to analyzing suspicious text, finding the employment of linguistic steganography, blocking, destroying and recovering the hidden secret information. Linguistic steganalysis is a research direction opposite to linguistic steganography and it includes two stages which are detecting analysis and recovering analysis. The research on linguistic steganalysis has an important significance in national security and social stability.In this dissertation, on one hand, we fully investigate the principle of various common linguistic steganography methods, work over the linguistic steganalysis problems, propose the linguistic steganalysis methods and carry out relative experiments. On the other hand, we design more secure linguistic steganography methods based on the understanding and analyzing the existent linguistic steganography methods.As a result, this dissertation focuses on the following research tasks.1) Research on the detecting analysis of common linguistic steganography methods: studying the blind-detecting idea of linguistic steganography, designing effective practical detection methods, programming corresponding tools and carrying out related experiments.2) Research on the recovering analysis of common linguistic steganography methods: reasonably defining and modeling the recovering analysis problem of linguistic steganography, studying the ideas of recovering analysis, designing corresponding recovering methods.3) Designing a system for text information hiding analysis, implementing the blind or nearly blind detecting analysis, in the view of system users, of text information hiding methods including all kinds of linguistic steganography methods.4) Presenting more secure linguistic steganography methods on the basis of learning from the existent linguistic steganography methods, and then theoretically analyzing and experimentally validating their security.According to the research tasks described above, this dissertation makes a variety of natural language processing techniques as the basic means, uses probability and statistics analysis as its main method and combines the idea of machine learning and artificial intelligence to do the research. Additionally, certain necessary theoretical analysis and experimental validation are carried out. Based on the research work described, this dissertation makes some innovative achievements including the followings.1) Aiming at testing text segments of different sizes, three blind-detecting analisys methods for linguistic steganography based on syntax, that is blind-detecting analysis method based on statistical characteristics of correlations between words, based on word position distributions and based on detection entropy, are designed, corresponding theretically analysis and experimental validations are done and these detection analysyis methods turn out to be promising. In addition to this, the recovering analysis problem of linguistic steganography method based on sentence template is defined and a feasible recovering analysis method is designed.2) An steganalysis scheme for linguistic steganography based on semantics is presented. A detection analysis method for linguistic steganography based on synonym substitution is designed and the experiment shows that the detection method is promising. Furthermore, a recovering method for linguistic steganography based on synonym substitution is proposed and related theoretical analysis is made.3) A method is proposed that we can realize blind or nearly blind detecting analysis of text information hiding through constructing a proper system. The properties that the system should have are expandability, adaptability, capability of feedback and learning. Based on the conception, a detailed system design scheme is presented and discussions are made to make it clear that how this scheme satisfies the required blind detecting analysis well.4) The security requirements of linguistic steganography are discussed and according to these requirements, two secure linguistic steganography methods, which are linguistic steganography based on double text segments and based on context synonym substitution, are presented. Finally, analysis and evaluation of security of these methods are made.In the achievements described above, points 1) and 2) improve the accuracy of linguistic steganography detection, enrich the detection technique of linguistic steganography and firstly discuss the problem of linguistic steganography recovering analysis. Point 3) proposes the method that realizes blind or nearly blind detection of text information hiding through constructing a proper system and a concrete system design scheme is presented. Point 4) brings forward two new and more secure linguistic steganography methods.