Study On Adaptive Video Steganography

Video steganography technique is an important branch of the technology of data hiding.Leveraging the technology of video steganography can embed secret messages into cover videos and generate stego videos.Then,the behavior of transmitting secret messages can be covered by transmitting stego videos,which guarantees the secure communication of secret messages.At present,the technology of video coding leads to the distortion produced by stego modifications to spread during the procedure of encoding videos,which results in the problem of distortion drift.Distortion drift reduces the visual quality and the security of stego videos.Therefore,this thesis mainly focuses on adaptive video steganography for decreasing the visual distortion,reducing the bit-rate increase,improving the embedding capacity and enhancing the security of stego videos.Four video steganography methods are proposed for Quantization Discrete Cosine Transform(QDCT)coefficients from the perspectives of reducing and exploiting distortion drift.The main contributions and innovations in this thesis are summarized as follows:1.A video steganography method is proposed by reducing the impact of distortion drift.There are two ways to reduce the impact of distortion drift.One is selecting special QDCT coefficients,whose modifications cannot cause significant distortions.Therefore,the proposed video steganography method selects middle and high frequency QDCT coefficients as cover elements.The other is reducing the number of modified QDCT coefficients.So the proposed video steganography method chooses EMD(Exploiting Modification Direction)algorithm for embedding and extracting of secret messages.Experimental results demonstrates that exploiting the two ways can effectively keep the visual quality of stego videos.2.A video steganography method is proposed by limiting the impact of distortion drift.Stopping distortions propagating and reducing the number of modified QDCT coefficients are two ways to limit the impact of distortion drift.The proposed video steganography method analyzes the intra-frame prediction modes of 4 × 4 transform blocks and then selects QDCT coefficients in 4 × 4 transform blocks with small impact of distortion drift as cover elements for stopping distortion propagating.In addition,the proposed video steganography method selects matrix embedding algorithm for embedding and extracting of secret messages,meanwhile reducing the number of modified QDCT coefficients.Experimental results demonstrates that Stopping distortions propagating and reducing the number of modified QDCT coefficients effectively improve the visual quality of stego videos.3.An adaptive video steganography method based on additive embedding distortion costs is proposed.The technology of adaptive video steganography has excellent performances in the visual quality and the security of stego videos.Therefore,designing additive embedding distortion cost functions reflecting the impact of embedding modifications plays key roles in improving the visual quality and the security of stego videos.Distortion drift reflects the impact of stego modifications,so the proposed method designs an additive embedding distortion cost function from distortion drift point of view.Experimental results demonstrated that designing additive embedding distortion cost functions from distortion drift point of view improves the visual quality and the security of stego videos.4.An adaptive video steganography method based non-additive embedding distortion costs is proposed.Non-additive embedding distortion costs consider the mutual impact between adjacent cover elements and improve the security of stego videos.Therefore,the proposed video steganography method divides a video into many non-overlapping blocks and designs joint distortion costs for QDCT coefficients of these blocks.Joint distortion costs reflect the mutual impact of changing inner-block QDCT coefficients of these blocks.Distortion drift reflects the impact of embedding modifications,so the proposed method designs a non-additive embedding distortion cost function from the perspective of distortion drift and distortion drift production.Experimental results demonstrated that designing non-additive embedding distortion cost functions from the perspective of distortion drift and distortion drift production improves the visual quality and the security of stego videos.