Inter-frame Correlation Based Compressed Video Steganalysis

A novel blind compressed video stream steganalysis algorithm employing the inter-frame correlation (IFCS) is proposed. The steganalysis process is executed wholly in the compressed domain for the real time requirement. Statistical features of inter-frame correlation, AC coefficients and DC coefficients in DCT domain of compressed video are extracted out as the classifier features using an inter-frame collusion principle. The collusion window is adjusted according to the inter-frame correlation. And one three layers feedforward neural network capable of nonlinear feature mapping is employed as the blind classifier. Experimental results verify the availability of the proposed scheme.