Steganalysis of Youtube Compressed Video Using High-order Statistics in 3D DCT Domain

This papers presents a novel steganalysis scheme for YouTube compressed video. The proposed scheme exploits spatial-temporal correlation between adjacent frames. To this end, 3D Discrete Cosine Transform (DCT) is used to capture this correlation. Statistical features sensitive to data hiding such as absolute central moments, skewness, kurtosis, and Markov features are used as classifying statistics. Unsupervised K-means clustering is used to distinguish between the cover- and the stego-videos. Effectiveness of the proposed scheme is tested using 40 standard videos. Experimental results reveal that the proposed scheme is capable of detecting the presence of hidden message in the stego-YouTube videos.