Research On Models And Methods Of 3D Steganography

In recent years,with the development of information technology,large amount of sensitive data including military,political,financial and commercial data are circu-lating on the Internet,thus,the security of these digital information is receiving un-precedented attention within countries,enterprises,or individuals.In order to ensure the security of digital information,necessary protective measures should be consid-ered.Digital steganography is a type of information hiding method utilized for covert communication and covert storage.Its main goal is to hide the secret information in digital carriers(such as images,videos,audios,3D models,text,etc.)and transmit them to the receiver without being observed by a detector.In the face of complex and diversified steganalysts,how to enhance the security of steganography has become an important problem of current steganography research.However,compared with image steganography,3D steganography still remains in the preliminary research stage and many key problems deserve investigation.As new data carriers,the 3D steganography model(3D meshes,3D textures,depth images,etc.)studied in this project has created a more secure way of covert communication,which has better resistance against potential steganalysis.Therefore,it is meaningful and valuable to study effective methods of 3D steganography.To improve the security of 3D steganography,three key problems need to be solved:3D mesh steganography,3D texture steganography and RGBD image steganography.Focusing on these three key issues,this dissertation investigates corresponding 3D steganography methods.The main innovations of this dissertation are listed as follows:1.Research on 3D mesh steganography·Security analysis of 3D mesh steganographyThe state-of-the-art mesh steganalysis methods extract features from ver-tices and edges,which are not effective in discriminating cover meshes and stego meshes.The dissertation proposed normal voting tensor based fea-tures to boost 3D mesh steganalysis,which forms a new security evaluation method for 3D mesh steganography.·3D mesh adaptive steganographyThe majority of existing 3D mesh steganography methods modulate vertex coordinates to embed messages in a nonadaptive way.The dissertation took account of complexity of local regions as joint distortion of a triple unit(vertice)and coding method such as syndrome trellis codes to adaptively embed messages,which owns stronger security.2.Research on 3D texture steganography·Security analysis of texture steganographyThe dissertation proposed a method to attack classic texture image synthesis based steganography methods.We find that the mirror operation over the image boundary is flawed and is easy to attack.The attack can not only de-tect the stego-images but can also extract the hidden messages.The disser-tation proposed another solution to improve the texture image steganalysis.By exploiting the observation that steganography destroys optimization of matching extent between the synthetic patch and optimal candidate patch,we reconstruct the two patches from an overlapped region to extract the existence of optimality,to boost texture steganalysis,which forms a new security evaluation method for texture steganography.·3D texture mapping based steganographyThe dissertation proposed a security-enhanced texture synthesis based stegano-graphic method by padding redundant regions carrying no message around the periphery of the synthesized image and generating additional candidate patches to increase capacity.To reach the goal of multi-domain steganogra-phy,the stego texture images are mapped to the stego 3D meshes,and they are considered as textured 3D meshes.3.Research on 3D depth image steganographyRGBD images are usually stored as two separate files in the same file path and utilized by users simultaneously.There are cases that the depth file may be lost during transferring or loading.Therefore,the dissertation proposed a novel algo-rithm for RGBD image steganography based on convolutional deep networks,us-ing two encoders to extract features and a encoder to encode the cascaded features into synthesized images.The method is a special application of steganography,that is,multi-modal integrated communication.