Steganalysis of LSB Speech with Low Embedding Rates based on Joint Probability
Wanxia Yang, Shanyu Tang · 2017
Steganalysis with low embedding rates is a difficult problem in the information hiding field. Using the characteristics of wavelet package decomposition, which is a delicate method for signal processing, and the feature of Joint Probability (JP), which better expresses the correlation of speech signals, a steganalysis at low embedding rates based on the JP feature of the second-order derivative-based Wavelet package Coefficient(WPC) of the speech signal is proposed in this paper. The steganography detection performance,including the detecting accuracy, computing complexity and extraction time of JP, was compared using Markov bidirectional transition probability (MBTP) and Traditional Markov second-order transition probability(TMSOTP) based on WPC and non-WPC, respectively. The experimental results indicate that the JP of the WPC is more suitable forsteganalysis at low embedding rates with the low complexity and shortest extraction time among the tested methods; the accuracy rate can be up to 75%,whereas the embedding is only 3%.