Research On Secure Reversible Data Hiding Algorithm In Encrypted JPEG Image Based On Coefficient Feature Change

With the rapid development of cloud platform storage technology,users choose to upload local images to the cloud platform for saving local storage space.Reversible data hiding in encrypted images(RDH-EI)is of great significance for cloud platform storage.Encryption makes the original image unreadable by cloud platform administrators,ensuring user privacy.RDH facilitates cloud platform management by embedding tag information.JPEG is a common image compression format,so RDH-EI in JPEG images has received extensive attention from researchers.In order to protect the user’s privacy,the security of the RDH-EI in JPEG image algorithm is very important.The main work of this thesis is as follows:According to the coefficient characteristics whether changes,the security of the existing RDH-EI in JPEG images is analyzed.It is pointed out that the image blocks in Qian have the characteristic that the ordered AC coefficients remain unchanged before and after encryption.And based on this,Qian algorithm is attacked.In known-plaintext attack,the partial block scrambling sequence is estimated by using the feature of keeping the ordered AC coefficients of blocks unchanged,and the corresponding AC coefficients are recovered.In the chosen plaintext attack,we construct a plaintext image with same image block bit stream and a plaintext image with different ordered AC coefficients of all image blocks.Then let them through the encryption system,the XOR cipher stream and block scrambling sequence can be estimated.The DC and AC coefficients of the image are recovered completely.In the end,sum up the security risks in existing RDH-EI algorithm in JPEG image:(1)The same key is used to encrypt multiple images,and there is an attack risk of known plaintext.(2)Even if the feature of non-zero AC coefficients is used to generate an adaptive key to resist knownplaintext attacks,but it is not changed during encryption and information hiding,there is still a risk of chosen-plaintext attacks.By studying the embedding algorithm based on zero coefficients,a JPEG image RDH algorithm that can change coefficient characteristics is proposed,which can change the number of non-zero AC coefficients in JPEG image while embedding secret information.The block selection and frequency selection strategies of zero-coefficient embedding are mainly studied to reduce visual distortion and file growth.In frequency selection,the effect of zero coefficient embedding on zero-run length of intermediate symbols is analyzed,and the simulation file increment of zero coefficient embedding at different frequencies is obtained.In block selection,a block smoothness function for block selection is defined as the weighted sum of two constant features,and the two constant feature selection,processing and weight selection strategies for zero-coefficient embedding are analyzed.Experimental results show that zero-coefficient embedding can change the coefficient characteristics effectively.Compared with non-zero embedding algorithm,embedding capacity of zero-coefficient can be increased by 8-17 times,and visual quality of the algorithm is improved by 1.2-2.5d B on average,especially for small quality factor.Compared with the existing zero-coefficient embedding algorithm,the unit increment of file size is decreased by 0.22～0.36,the visual quality is improved by 1.8～2.3d B on average.Combining encryption algorithm based on adaptive key and RDH algorithm with zero coefficients embedded to change the characteristics of the coefficients,a more secure RDHEI algorithm in JPEG images based on the coefficient characteristics change is designed.In terms of algorithm security,using the feature of non-zero AC coefficients to generate the encryption key can resist known plaintext attacks;using the zero-coefficient embedding algorithm,the number of non-zero AC coefficients is changed,which can resist the exist chosen plaintext attack.In terms of embedding capacity,maximum embedding capacity of can reach 33 times that of the zero-run value rotation embedding and 6～14 times to the non-zero embedding.The algorithm can achieve a good balance between embedding capacity,file growth and algorithm security.