Audio steganalysis based on factor analysis and support vector machine
Jinsha Yuan, Yincheng Qi · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2008
Factor analysis (FA) is a multivariate statistical method for studying the correlation among a group of variables, and is able to minimize data redundancy and reveal hidden patterns. We used FA to reduce the dimension of feature vector and support vector machine to classify the audio data sets in the proposed audio steganalysis system. Then the system is utilized to detect the stego-audio signals embedded by wavelet domain LSB, quantization index method (QIM) and addition method (AM). The dimension of feature vector is decreased from 36 to 7 (LSB), 6 (QIM), 6 (AM) respectively. The detection rates are all greater than 91% and the detecting performance is maintained.