Research And Design Of Video Steganography Analysis Method Based On Motion Vector

Steganography technology plays an important role in ensuring the security of information and communication in military and intelligence fields.However,the development of steganography technology also poses a threat to national security.Terrorists and illegal organizations use this technology to plan illegal activities,which seriously threatens social stability and national security.Therefore,the research of steganalysis technology is of great significance.Steganalysis technology can be divided into image-based steganalysis,audio-based steganalysis and video-based steganalysis.Because video as a steganographic carrier has more advantages than image and audio,steganographic analysis technology for video has attracted wide attention.Steganalysis based on motion vector is one of the hotspots in video steganalysis.How to improve the detection accuracy of steganalysis is an important direction in the research of video steganalysis technology.This paper improves the detection accuracy by extracting more effective features and using better classifiers.The main research work and innovations are as follows:1.A video motion vector steganalysis method based on code generation value change is proposed.The re-encoding sub-pixel encoding cost can be well matched to the encoding cost before re-encoding.The 11-dimensional MVCCC(Motion Vector Coding Cost Change)feature is formed by statistically changing the sub-pixel coding cost corresponding to the motion vector before and after the video re-encoding.The experimental results show that the feature is about 10%higher than the typical IMVRB feature and AoSO feature detection accuracy.2.A video motion vector steganalysis method based on local optimal probability variation of coding cost is proposed.Based on the SPOM feature,the change of the local optimal probability of the coding cost is added,and the range is extended to form a 12-dimensional SPCCOM(Subtractive Probability of Coding cost Optimal Matching)feature.The experimental results show that the feature is about 12%higher than the typical SPOM feature detection accuracy.