Abstract: A new steganalysis scheme based on taking the inter-frame correlation as the undirected graph weights was proposed utilizing the temporal correlation among video frames. First, getting the eigenvector of every video frame brightness by calculating its gray-level co-occurrence matrix, the eigenvector is an eight-dimensional vector. Then, computing the weights of each video frame by using the euclide distance algorithm. The weighted undirected graph which representing the correlations between each frame is constructed. Finally, according to the change of correlations between the frames after embedding information determine whether the video is embedded confidential information. Experimental results show that the method using weighted undirected graph can quickly discriminate stego video and the original video, and have a high accuracy rate.