Steganalysis on positive and negative echo hiding based on skewness and kurtosis

Positive and negative echo hiding is a robust and imperceptible data hiding method. In this paper, an effective steganalysis algorithm based on skewness and kurtosis is proposed. According to echo in stego speech, both power cepstrum and absolute value of complex cepstrum of speech are analyzed. For each cepstrum, its high order statistics including coefficients of skewness and kurtosis are calculated in combination. By training with SVM the combinatorial skewness and kurtosis in each cepstrum of clean speech and stego speech, we obtain combinatorial skewness-kurtosis thresholds, which can be used for blind steganalysis of echo hiding. Experimental results show that our proposed steganalysis does well in detecting positive and negative echo hiding, and absolute value of complex cepstrum performs better than power cepstrum.