Research On Asymmetric Steganographic Distortion Model And Optimization Methods For Images

The wide application of information technology and the development of cyberspace greatly promote the prosperity and progress of economy and society,but it also brings new security risks and challenges.Confidential information such as commercial,military and personal privacy is spread in cyberspace and faces the risk of being stolen and leaked.How to ensure the security of the transmission of these confidential data in cyberspace has become a hot issue in the field of information security.Traditional solutions are based on cryptography,however sending and receiving cipher messages is prone to raise suspicion,thus leading to the exposure of secret communication.In an uncontrolled environment,confidential communications need to protect not only the content of the message,but also the fact of communication.Communications that satisfy these two requirements are also known as covert communications.Steganography,as an important technique for covert communication,conceals the fact of secret communication,circumvents third-party sabotage from the perspective of behavioral security,and plays an indispensable role in safeguarding the communication security of special state departments and key groups.As the most common media type,digital image steganography has become a hotspot.However,there are still many problems and challenges in the application of image steganography in real scenes.Firstly,the images propagated in cyberspace are mainly color images.However,steganography researches for images mainly focus on grayscale images,which cannot meet the demand for color image steganography;Secondly,most of the social platforms in cyberspace will lossy compress the uploaded images,thus destroying the hidden messages.Therefore,it is a challenge for steganography to balance robustness and anti-detection;Finally,traditional image steganography usually uses symmetric distortion,which assigns the same distortion to the+1 and-1 modifications for each image element.However,symmetric distortion ignores the correlation between the modification patterns of image elements and the distribution of elements,which limits the anti-detection and robustness of steganography.Therefore,it is an important issue to define asymmetric steganographic distortion for color image and JPEG image to effectively improve the anti-detection and robustness of steganography.To address the above problems and challenges,this dissertation designs asymmetric steganographic distortion models for color images and JPEG images respectively,and designs optimization methods based on these models to improve the anti-detection of steganography.For lossy processing on social platforms,this dissertation designs asymmetric steganographic distortion models and optimization methods to resist JPEG compression,and improved the anti-detection and robustness of JPEG image steganography under lossy compression environment.The main work and innovations of this dissertation are summarized as follows:1.Asymmetric steganographic distortion model and optimization method for color imagesIn order to maintain the correlation and difference between the RGB three channels in steganographic modifications to improve the anti-detection.In terms of correlation,this dissertation experimentally verifies that the R and B channels have a strong correlation with the G channel,and encouraging them to synchronize with the modification direction of the G channel in steganography can achieve better anti-detection.In terms of difference,this dissertation points out that the embedding capacity of the three channels differs greatly and the distribution of complex regions among the three channels is also different.Accordingly,this dissertation designs an asymmetric steganographic distortion model for color images and proposes the G-channel-related inter-channel asymmetric steganographic distortion optimization method GINA(G-channel-related Interchannel Non-Additive),based on which the asymmetric distortion can be defined to encourage the R and B channels to have the same modification direction as the G channel during embedding,and the payload can be distributed adaptively between the three channels while maintaining the original distortion distribution.The experimental results show that the proposed method can better resist color image steganalysis.2.Asymmetric steganographic distortion model and optimization methods for JPEG imagesBy comparing the detection resistance between the block interiors and block boundaries in the spatial domain after JPEG steganography,it is found that the modifications of the block boundaries are more easily detected by steganalysis.Therefore,it is necessary to enhance the detection resistance of block boundaries in the spatial domain in steganography.Firstly,this dissertation proposes the principle of Block Boundary Maintenance(BBM)from the perspective of maintaining the distribution of modifications in the block boundaries of the spatial domain.Based on the BBM,an asymmetric steganographic distortion model using the intracorrelation of coefficients in the DCT block is designed,and an updating method is used to reduce the modification of the block boundaries of the spatial domain,thus enhancing the detection resistance of JPEG image steganography.Secondly,this dissertation designs an asymmetric steganographic distortion model using all inter-correlations between DCT blocks from the perspective of maintaining the Block Boundary Continuity(BBC).Based on the BBC,the updating of covers and distortions is used to better maintain the continuity of modifications between the block boundaries of the spatial domain,which effectively enhances the detection resistance of JPEG image steganography.Finally,this dissertation combines these two principles to design an asymmetric steganographic distortion definition method that exploits both intra and inter-block coefficient correlation,thus further improving the detection resistance of JPEG image steganography.3.Asymmetric steganographic distortion model and optimization method for JPEG compression resistanceThe commonly used social platforms in cyberspace apply JPEG compression to the uploaded images.However,traditional adaptive steganography cannot correctly extract secret information from the JPEG-compressed cover image because it does not take robustness into account.For this reason,robust steganography is proposed to improve the robustness of steganography at the expense of partial resistance to detection.In order to improve the robustness and detection resistance of JPEG image steganography at the same time,this dissertation explores the robustness of different DCT coefficients in JPEG images under JPEG compression and the anti-detection performance after different modifications of DCT coefficients,and finds that reducing the absolute value of DCT coefficients can improve the robustness and anti-detection performance of steganography at the same time.Accordingly,an asymmetric steganographic distortion model with resistance to JPEG compression is designed,and a method for defining asymmetric steganographic distortion is devised.The experimental results show that the proposed asymmetric steganographic distortion can effectively improve the detection resistance and robustness under various robust steganography frameworks,and can further improve the detection resistance of adaptive steganography in lossless channels.