A Video Steganalysis Algorithm for H.264/AVC Based on the Markov Features

A new steganalysis algorithm based on the features of the original domain was proposed combining spatial correlation with temporal correlation among video frames. First, divide video frames into a new kind of special frame structure of P0-I-P1, calculate the luminance component for every frame of the macroblocks. And then we use the Eulerian-distance of vectors to express the correlation between macroblocks to get the most similar macroblock of adjacent frames. Thus, we combine the co-occurrence matrix model with Markov model into the P0-I and I-P1 frame type. We try to determine whether any data hiding is embedded in the video according to the changes of correlation between video frames. The experimental results show that the algorithm proposed in this paper has a high detection probability, and it calls the promising result and deserves further studies.