Research On Steganalysis In Midi Audio

Steganography is an important branche of information hiding, and a primary technique instrument that is used to actualize covert communication. Steganalysis is attack to steganography. Its purpose is to examine the existence of secret information up to destroy covert communication. In recent years, as carriers of information hiding become diversification, MIDI is used to be the carrier of the secret documents. In order to protect the security of information networks, put forward a dedicated MIDI audio steganalysis algorithm is of great significance. Since the MIDI format of the store is document, the structure is different with wav audio. LSB is the most commonly method used to hide information in MIDI audio, and people choose the parameter music velocity to hide information. At the same time, due to the specificity of MIDI audio,we can hide information in MIDI audio based on the characteristics of format.Paper studies steganalysis in MIDI audio based on the changes in the demographic characteristics. The main work and innovation is as follows:1. Applies the typical Steganalysis methods in images to MIDI audio, studies RS analysis, GPC analysis and entropy analysis, compare these methords on the effect of detection through experiments. Experiments show that when embedding rate is high, entropy analysis can effectively steganalysis.2. In allusion to the LSB steganography statistical characteristics velocity in MIDI audio, put forward a dedicated MIDI audio steganalysis algorithm using the idea of difference histogram. Compare to the typical Steganalysis methods in images,our methord has lower FPR and higher TPR, even when embedding rate is very slow, it still has good results. Through the use of SVM classifier, the recognition rate of MIDI audio is up to 97.5%.3. In allusion to the steganography based on the characteristics of format, by dealing with MIDI event, successfully detect the steganography with the idea of whether whrite or not the same command next to each other. Statistics show that MIDI audio strictly follows the principle of the smallest, and omits the same command next to each other. So steganalysis algorithm's FPR is close to zero and TPR is close to one. Algorithm can be applied to steganalysis in MIDI audio with the idea of whether whrite or not the same command next to each other on the Internet.