Audio Steganalysis Based on Co-occurrence Matrix and PCA
Yinchen Qi, Yan Wang, Jinsha Yuan · 2009
A new steganalysis scheme based on co-occurrence matrix for audio signals is proposed. The statistics features are derived from the co-occurrence matrix firstly, which are calculated from amplitude of audio signals. Then the preprocessing of principal component analysis (PCA) is used on statistics features and the support vector machine (SVM) is used as a classifier. Experiment results for 450 audio signals of CASIA98-99 audio database show that the detection rates of three audio data hidden methods (wavelet domain least significant bit, quantization index method and addition method) are all greater than 92%.