| Digital steganography and steganalysis are important branches of information security.Because of the large amount of data and the ability to accommodate a large number of secret information,digital video has become an ideal steganographic carrier.As the latest video coding standard,H.265/HEVC has been gradually applied to a variety of products in the Internet.The study of H.265/HEVC based steganography and steganalysis algorithms has important theoretical significance and practical value.The motion vector,as introduced by video inter prediction,is widely found in compressed video.Motion vectors provide a large amount of data and have very limited influence on the visual quality after message embedding,making them very suitable for information hiding.However,these features of motion vector are likely to be used for illegal purpose such as passing secret information to plan criminal activities.Therefore,the research on video steganalysis based on motion vector has an urgent demand and broad application prospects.To address the shortcomings of the existing motion vector based steganalysis algorithm,we employed the characteristics of H.265/HEVC to construct more efficient features for steganalysis detection.The main contributions of this work are as follows:1.We proposed an H.265/HEVC video steganalysis algorithm based on optimal predictive motion vector.The algorithm utilizes the property that the predictive motion vector is the optimal one(i.e.the one minimizing the number of coding bits of the current motion vector)in the candidate motion vector array during the H.265/HEVC motion estimation process.The experimental results show that the proposed algorithm has higher detection rate than the existing steganalysis algorithm based on motion vector correlation and the steganalysis algorithm based on local optimality of motion vector.Moreover,as the motion vector is lossless,the detection result is not affected by the video bitrate.2.We proposed an H.265/HEVC video steganalysis algorithm based on adaptive selection of temporal and spatial domain features.The existing steganalysis based on motion vector correlation only extracts the spatial domain(or time domain)correlation feature or simply concatenates the spatial domain correlation feature and the time domain correlation feature.The proposed algorithm utilizes the characteristics of H.265 / HEVC motion vector prediction and adaptively select spatial or temporal correlation characteristics to form the final classification characteristics.The experimental results show that the proposed algorithm can improve the detection rate of steganalysis without increasing the feature dimension. |