MIDI Audio Steganography And Steganalysis

Steganography and steganalysis are two main areas of Information hiding technology. Steganography is to hide secret information into the carrier. Steganalysis purposed is to detect and destroy the secret information is an attack on steganography. In recent years, with the development of information hiding technology, MIDI audio is being the carrier hidden secret message. MIDI file with the advantage of small size and editing easy is stored on the instructions of how to play music. MIDI is a music information transmission protocol standard between Music equipment and computers, which is widely used in the field of mobile phones and the internet. As a result, MIDI audio information hiding technique is more important.MIDI audio is stored as MIDI commands. There are both similarities and differences between MIDI file information hiding and other carieers. This paper mainly research on MIDI audio information hiding supported by National Natural Science Foundation, thesis work and innovations are as follows:1 Summary audio steganography and steganalysis method, proposed a more complete audio steganography and steganalysis evaluation model. Summarize the MIDI audio steganographic methods and steganalysis methods according to the evaluation model.2 As there is only MIDI audio LSB replacement Steganalysis method, a method for MIDI audio LSB matching is proposed in the paper. The method judge a MIDI audio file by training smoothness conversion rate threshold, as the process of Steganography would change velocity smoothness of the MIDI file. The method can return a probable embedding rate based on additive noise feature of the steganography. Experiments show that when the embedding rate greater than 20%, the correct detection rate can reach more than 70%, also the error of the embedding rate is less than 10%.3 In order to improve MIDI audio LSB detection efficiency, a steganalysis algorithm is proposed based on the characteristic mining and pattern recognition. Extract 21-dimensional statistical characteristics of histogram characteristic function (HCF) field, training classifier with the support vector machine (SVM). Experiments show that the average classification accuracy rate can be above 90% for either LSB replacement or matching steganography. Finally, propose a LSB steganalysis optimization feature extraction methods using the characteristic of MIDI audio sub-channel storage. Experiments show that both classification accuracy and applicability increased.4 A MIDI blind steganalysis classifier was designed by Analyzing MIDI files structural features and statistical features for MIDI audio LSB steganography, omitted instructions steganography and synchronization instructions steganography. Experiments show that the blind classifier can detect three methods steganography of MIDI audio effectively. Finally, design a simple online steganography and steganalysis system using the browser and server (B/S) mode, which can achieve LSB steganography and Steganalysis of MIDI audio online.