MIDI Audio Steganalysis via HCF-based Statistical Features
Cuiping Wang · Communications technology · 2010
In view of 3 kinds of LSB Steganography for MIDI audio,that is,LSB replacement,LSB matching and 2LSBs replacement,a method of steganalysis is proposed,in order to improve steganalysis accuracy based on the statistical features of the histogram characteristic function and SVM classifier.The 21-dimensional statistical features of the histogram characteristic function are extracted to classify original MIDI audio and stego MIDI audio.Experiments show that when embedded rate is more than 10%,steganalysis by the proposed method could detect MIDI audio with an average correct decision rate of above 90%.