Research Of Asymmetric Distortion Steganography Algorithm Based On Ternary Embedding

Steganography is an important branch in the field of information hiding.Its purpose is to embed secret messages in digital carriers and make the secret carriers difficult to be detected in the public network channel transmission.At the same time,the wide application of digital image in the era of big data makes it become the most favorable communication carrier in steganography.However,in order to reveal the behavior of secret messages in secret carrier images,steganalysis technology poses a threat to the security of existing steganography methods through feature extraction and training of massive image data.Therefore,it is necessary to study new steganalysis methods to ensure the security of information transmission.This paper proposes an improved model for ternary embedding steganography in the field of digital image steganography,and optimizes the distortion cost function currently used.The specific work is as follows:(1)A ternary asymmetric distortion model for spatial steganography is proposed.The model constructs a reference image which is closer to the natural image through the original image,and uses the reference image to guide the embedding modification direction at different positions.The existing distortion cost function is optimized by the model,and different adaptive distortion values are set during the +1 and-1 operations at the embedding positions.At the same time,the generated secret carrier is closer to the constructed reference image,so as to improve the security of the secret carrier.Experimental results show that this method has better performance against steganography detection.(2)A smoothing distortion algorithm for spatial steganography is proposed.In order to maintain the content relevance of the original carrier image in the steganography embedding process,this method blocks the image and embeds it layer by layer iteratively.The influence of the distortion of the embedded layer on the neighborhood guides the modification of the non-embedded layer,so that the modification direction in the neighborhood is as consistent as possible,so as to maintain the content relevance of the secret image.This method uses the steganographic image embedded in each layer as the intermediate carrier,which smoothes the distortion value in the neighborhood and strengthens the correlation in the neighborhood.Experimental results show that this method can greatly improve the security of existing methods.(3)A fast steganography method for Joint Photographic Experts Group(JPEG)images is proposed.In this method,Kullback-Leibler(KL)divergence is used to measure the optimal risk of the distribution of vectors before and after steganography,and the Fisher information matrix of the image is used to ensure the security of the steganographic image.The Laplace distribution is used as a model to fit the distribution of Discrete Cosine Transform(DCT)coefficients of JPEG images,which greatly improves the calculation speed of distortion cost function and the embedding speed of steganography.The experimental results show that this method can achieve millisecond embedding in JPEG images,and has certain anti-detection.