A detection method of subliminal channel based on VoIP communication

Abstract

With VoIP (Voice over IP) is widely applied in the Internet, Using the subliminal channel based on VoIP to transmit secret message has been a significant issue in Information Security. At present, many kinds of information hiding algorithms have been designed and applied in real-time VoIP communication, and the most widely used algorithm among them is the LSB (Least Significant Bit) hiding. However, it is difficult to detect the LSB matching steganography in real time. In response to this challenge, this paper raised a structure of system, which is able to capture the VoIP streams on the Internet and detect the real-time subliminal channel in them. We also designed a new steganalysis algorithm for the system which is aimed at the LSB matching steganography algorithm. It used the feature extraction method based on correlation and SVM (Support Vector Machine) classification. The experiment results demonstrated that while the embedding rate was over 50%, the new algorithm raised the accuracy of detecting the LSB matching by about 20% compared with existing algorithms. At the same time, the length of the samples cannot obviously affect the accuracy of the new algorithm; it can detect the information hiding in the samples whose length is only 3s. From this point of view, the new algorithm has good real-time performance.