It is well established that using the selection channel, the probabilities with which the elements in cover are modified during message embedding, would improve the performance of steganalysis. Most video steganographical algorithms embed secret messages in the compressed domain by modifying the motion vectors having less impact on video visual quality, which can be considered as a form of selection channel. Recently, deep neural networks have been rapidly developed for multimedia steganalysis. Although there have been some selection-channel-aware networks for image steganalysis, they cannot be simply extended to video steganalysis because there are great differences between image and video steganographic modification. To our best knowledge, there have been no selection-channel-aware networks for video steganalysis in literature. In this article, we propose a selection-channel-aware deep neural network for video steganalysis. Considering that video structure is quite different from that of image, we focus on the construction of input data matrix for deep convolutional neural network, the definition of probability for motion vector modification, and the network structure of using the selection channel knowledge. Experimental results have demonstrated that the proposed method benefits from selection channel and has satisfactory performance on testing HEVC videos.