Research On Key Issues Of JPEG Image Steganography

Digital steganography embeds secret information into digital images,audio,video,and other media,and transmits them through public channels to achieve covert communications.It has wide application prospects in military,security,personal privacy and other fields.In recent years,the rapid development and widespread application of steganalysis technology and smart mobile devices have brought new challenges to digital image steganography research.How to design a digital steganography algorithm that is safer,more efficient,and adaptable to new open channels is the key issues of JPEG steganography.This paper begins with a public question on steganographic detection feature selection,and focuses on how to improve the anti-detection and anti-JPEG compression of JPEG image steganography algorithms.The contributions are as follows:1.On problem of feature selection between probability density function?PDF?moment and characteristic function?CF?moment.The detection performance of the PDF moments and CF moments are analyzed theoretically in the logarithmic prediction error?LPE?wavelet subband.First,the specific distribution of coefficients on LPE wavelet subband is deduced based on the definition of Gaussian distribution and logarithmic prediction error subbands.Then,the four lemmas and corresponding proofs of the PDF moments and CF moments satisfying this distribution are given.Last,the detection performance comparisions of the first-order and second-order PDF and CF moments of the LPE wavelet subband are proved based on these lemmas:The first-order PDF moment is better than the first-order CF moment while the second-order CF moment is better than the second-order PDF moment on the logarithmic prediction error wavelet subband.It clarifies the experimental inference of Moulin et al.published in IEEE TIFS papers in 2007.A large number of experiments were performed based on the BOSSbase image library.The results verified the reliability of the above theoretical results.2.To elevate the embedding efficiency and detection resistance in the existing binary embedded edge information JPEG image adaptive steganography,an±1 adaptive JPEG steganography method based on adaptive payload two dimensional decomposition embedding method is proposed.First,the anti-detection analysis of edge information JPEG adaptive steganography algorithm is carried out,and the conclusion that ternary embedding is better than binary embedding is obtained.Second,the preliminary design of ternary embedding cost function is given,and pointed out that the ternary embedding cost function needs to be carefully designed,otherwise an unreasonable design may reduce the steganalysis resistance based on a large number of experiments.Third,a adaptive payload two-layer decomposition embedding method?2D-DEM?that can maintain the minimum distortion is proposed.Fourth,a universal design idea of the ternary embedded cost function is designed based on the 2D-DEM,and 3existing binary side-information JPEG adaptive steganographies are improved by it.Last,the parameter optimization setting strategy of the ternary embedding cost function is proposed based on the contribution 1.Compared with the existing three kinds of representative edge information JPEG adaptive steganography experiments,the results show that the proposed ternary adaptive JPEG steganography owns higher resistance to statistical detection.3.On the problem that the existing anti-JPEG compression robust steganography algorithm DCRAS can not extract the embedded information completely,a robust anti-JPEG compression steganography method with high extraction accuracy is proposed.First,the concept of a robust virtual carrier and a construction method are given based on transform domain relative coefficient relationship and lossy channel compression quality factor.Second,a robust steganography-based serial level is designed based on a robust steganography framework.At the same time,a concatenated error correction code is proposed by concatenating the convolutional code and the RS code through an interleaving cascade on the requirements of high code rate and continuous error correction performance.The analysis of the steganographic algorithm and the experimental results based on the BOSSbase image library show that,the embedded information extraction accuracy of the newly proposed anti-JPEG compression steganography is very close to 100%(error rate is about1.77?10^-22),and the anti-detection ability is better than DCRAS.4.On the problem of invalidation of existing anti-JPEG compression robust steganography in the condition without the side-information of lossy channel,a robust steganography method without side-information.First,the failure of information extraction caused by the error diffusion of STCs is analyzed.Second,a robust virtual carrier is constructed under the condition without side-information.In this method,the robustness of the robust virtual bearer is removed from the lossy channel side-information.Last,an error diffusion error control algorithm for STCs decoding is proposed to further ensure the correctness of information extraction.Comparative experiments on anti-JPEG compression performance and anti-statistical detection performance are conducted with the existing JPEG adaptive steganography,robust watermarking,and robust anti-JPEG compression steganography.The results show that the newly proposed anti-JPEG compression robust steganography can resist JPEG compression without lossy channel side information at the expense of a small amount of anti-statistic detection performance.5.In order to further improve the anti-JPEG compression robustness and detection resistance of the steganographic algorithm without the lossy channel side information,a novel robust steganography based on embedding region selection and cascade error correction code is proposed.First,based on Harris-Laplacian transform,the invariant feature points with strong anti-detection in the image are found out,and the setting method of the radius of the embedded region is given.Second,based on the generating method of robust virtual carrier without side information conditions,the high modification probabilities on low-security elements is eliminated in secret embedding by designing a more reasonable embedding cost function.Last,the information embedding is achieved by combining the concatenated error correction code and the STCs error diffusion control algorithm.A large number of experimental results based on the BOSSbase image library show that the proposed anti-JPEG compression robust steganography greatly improves the accuracy of secret information extraction without the lossy channel side information.At last,the full-text works are summarized and the next possible research directions are prospected.