Research On Algorithms And Practices Of Stegonagraphy And Steganalysis

Through more than 10 years' development, information hiding technology has become an important academic branch of information security. Battles between steganography and steganalysis have become an important issue in information hiding, which makes steganography and steganalysis restricts mutually, and promote mutually. This thesis is focused on the techniques of steganography and steganalysis in images and audios, and the practical applications of steganalysis techniques. Two steganalysis algorithms and one steganography algorithm are presented. Based on the study of practical applications of detection techniques, the detection and monitoring system against covered communication over networks is designed and implemented. The main content of this thesis is summarized as follows:1. A steganalysis method of LSB replacement in wav audio is proposed. The probability model of switch rule between sample values is constructed based on the analysis of the LSB (Least Significant Bit) replacement scheme in wav audio. By adopting the methods of Povs detection, RS detection and DIH detection, and making improvements to these methods, the reliable detection to LSB replacement in wav is achieved. Experimental results show that the three algorithms can detect LSB replacement in 8-bit and 16-bit coded wav audio reliablely. When the embedding ratio of these two coding manners is above 3% and 20% respectively, the correct rate of stego-audio's judgments can reach 100%.2. Audio signal has more information redundancy than image signal, but its randomicity is stronger. It is more challenge to extract embedded bit with non-error from audio. A novel audio steganography based on QIM (Quantization Index Modulation) in the DCT (Discrete Cosine Transform) domain is proposed. The QIM quantizer is constructed on the basis of QIM principle. The message is embedded and extracted by the mapping between message bit and the quantizer's quantization interval. The extraction is executed blindly. The robustness of the hidden message is ensured well by the improved QIM, which employs the error-tolerance operation during embedding and extracting. The experimental results show that the capacity of the proposed method is large as 357.6b/s and good imperceptivity is achieved. The embedded message has good robustness against MP3 compression, requantization, resampling and lowpass filtering.3. Based on the probability model of switch rule between pixel values in LSB matching scheme, we propose a new steganalysis technique which aims at LSB matching embedding based on the image's smoothness in local area. The detection method defines the correlation between regional pixels as the difference between the sum of 8-neighbor pixel values and the center pixel. The existence of embedded message can be detected by observing the change of the regional correlation before and after embedding. Compared with the existing algorithms, the advantages of the proposed method includes simplicity, computational efficiency and low false positive rate. The detection achieves better reliability when embedding rate is above 30%.4. Aiming at secret communication contained in multimedia message over Internet and MMS (Multimedia Message Service) networks, the detection and monitoring system is designed and implemented on the basis of covered communication model over networks. Two instances of system's applications are illustrated in actual network communications environment with the help of typical steganography software. The instances explain the system's effectivity to monitor the secret message transfered over networks. Moreover, the system meets the real time requirement of MMS communication.5. The practical extraction of hidden information is proposed on the basis of the signatures of typical steganography tools. Many steganography tools may result in obvious identifications in stego-objects. By locating these identifications, not only the tools used to hide information can be found, but also it will be made sure where the hidden information is to those tools whose embedding rules are known. This makes the extraction of the hidden message become realizable, which is meaningful to the practicality of steganalysis techniques.