EMD(n,m) Model And Its Application In Digital Image Steganography

Image steganography hides secret information in unrelated carrier by using image redundancy and visual masking,which makes it difficult for illegal attackers to find the stego-image and extract the hidden secret information from mass and publicly transmitted images.Many traditional steganography methods base on LSB whitch ajusts a lot of carriers.Meanwhile,embedding rate by LSB is low.The EMD methods increase embedding rate by adjusting few carriers,But they only use the continuous figure combinations started by 0 as the expression range of secret information.There are only several limited kinds of EMD methods by using special basis vectors,such as simple EMD which adjusts at most 1 data in n carrier data,EMD-2 which adjusts at most 2 data in n carrier data,EMD-n,EMD-n±2,EMD-n±x and EMD-cn which adjust at most n data in n carrier data.these embedding methods limit not only expression range of secret information but also EMD's applications,and meanwhile decrease steganography security.To address these problems,the finished works are as follows:1)An EMD model EMD(n,m)based on free-basis vector was proposed.This model firstly formed the embedding adjustment table of secret information by calculating all m out of n carrier data combinations and then selected one row from the embedding adjustment table of secret information to embed the corresponding secret information.To further increase security,this paper also proposedan image steganography method based on EMD(n,m)model.First this method scanned each carrier into 1D sequence and randomly generated m out of n carrier data combinations to form the embedding adjustment table by user key and chaotic map.Second this method truncated all carrier sequence elementsand secret information bits respectively according to the carrier number and the number of secret information combinations.Finally secret information was embedded by adjusting carrier elements according to the mapped row by secret information in the scrambled embedding adjustment table.Theory and experiments show that by compared with traditional EMD methods,the EMD(n,m)model avoids the limited embedding capacity brought by the limited scope of the basisvector and it can maximize the application of the EMD method and also further enhance the embedding security by adding carrier and user key related the embedding-extraction procedures.2)For further improving the EMD(n,m)model's embedding capacity,the EMD(n,m)model is modified to EMD(n,m.?)model by extending EMD(n,m)model's modification from 1 to ?.To maintain the visual quality of the embedded carrier,an adaptive steganography method based on image block complexity and EMD(n,m,?)model was proposed.The proposed method firstly generates the maximum 8 dynamicly by PSNR and SSIM after the carrier image was modified randomly,then divided the image into several blocks,according to the 5 parameters:image block's intensity,contrast,location,edge and texture to jointly determine the image block complexity.The modifications in image blocks which have the lowest complexity is 1 and the highest complexity is ?,other image blocks whose complexity between the lowest and the highestwould be set between 1 and ? according to the logarithmic function randomly.To avoid the same embeding PSNR with the same resolution and the same used keys meanwhile to increase the visual mask,the n carrier data and the most m modified data would be decided by the user keys which is mapped by carrier's SHA256 sequences and the user key nmaX.The number of secret information combinations and the embedding adjustment table was generated by usingn,m,? jointly.According to the number of secret information combinations,2-value bits from secret information sequences was truncated and transformed into decimal number and the corresponding row fromthe embedding adjustment table was selected to embed secret bits in order of block complexity from high to low.Theory and experiments show,by compared with EMD(n,m)methods,the EMD(n,m,?)method can maximum embedding rate and achieve higher visual quality and structural similarity at the same time,meanwhile the procedure of embedding relies on the user keys and carrier's SHA256 sequences,and the secret information is embedded in order of image block complexity from high to low,all of this make the proposed methods with higher visual mask and enhance the embedding security.