Research On Side-Information Estimation For Image Steganography Method

Steganography is a major subfield of the field of information hiding,it is an important supplement to the technology of encryption,the message is first encrypted,then embedded into a cover,and then sent.Not only is the message’s content concealed,but also the communication process is kept secret.In recent years,there have been numerous studies on digital image steganography,in which side-information steganography uses the extra knowledge carried in the original image to adjust the initial modification distortion and significantly improve the detection resistance of steganography behavior,but it needs the original high-quality image for reference.In the case that high-quality images cannot be accessed,side-information can be obtained by estimating precover to provide auxiliary knowledge for distortion adjustment.This method is called sideinformation estimated steganography.Therefore,the steganography method of image side-information estimated is studied in this dissertation.Firstly,an side-information estimated steganography method based on super resolution network is proposed.Furthermore,an image steganography method combining real and estimated side-information is proposed.The chief studies are summarized as follows:1.Propose a method for the spatial image steganography of side-information estimated based on super-resolution networksIn real-world situations,steganographers frequently lack access to the process of collecting cover images,preventing them from employing high-security sideinformation steganography techniques.This dissertation proposes reconstructing high-resolution images using super-resolution networks,extracting estimated sideinformation during the downsampling process,and adjusting the modification of the distortion calculated by the underlying distortion function based on the polarity adjustment strategy,thereby reducing the overall distortion caused by the embedded information on the cover.In addition,in order to use two different estimated side-information simultaneously,this dissertation proposes a new distortion adjustment strategy that permits the steganographer to use both different estimated side-information in order to evade steganalysis detection.Experiment results demonstrated that the security of the sideinformation estimated method proposed in this dissertation for spatial domain images is not only significantly superior to the conventional method,but also superior to method based real side-information for high embedding rates.2.Propose a method of image steganography combining real and estimated side-informationThe results of Study 1 show that the security of side-information estimated steganography can exceed that of real side-information steganography,suggesting that the estimated side-information can obtain effective information other than real sideinformation.Inspired by this,a steganographic method combining real side-information and estimated side-information is proposed in this dissertation.For spatial image,the downsampling process of high-resolution image can be directly accessed to extract real side-information,and the estimated side-information can be obtained by super resolution method.Correspondingly,for JPEG images,the JPEG compression process of the original spatial image can be directly accessed to extract the real side-information,and the estimated side-information can be obtained by image deblocking method.In this dissertation,a new distortion adjustment strategy is designed to integrate real and estimated side-information.Experimental results show that the proposed method is safer than real side-information steganography in both spatial and JPEG images,proving that estimated side-information can provide knowledge other than real side-information,and the two complementary methods can generate steganographic distortion more resistant to detection.