Research On Coverless Information Hiding Based On Partial-duplicate Image Retrieval

In the traditional image information hiding method,the secret information is hidden by modifying the image,so it cannot resist the detection of the steganalysis tools effectively.However,the coverless information hiding can communicate secret information without any modification and can resist the detection of the steganalysis tools effectively.The coverless information hiding includes coverless information hiding based on text and coverless information hiding based on image and others.The existing coverless information based on image communicates secret message by using the relevance between the image and itself.And the redundancy of image information results of a low hiding capacity.Meanwhile,the security of coverless information hiding based on image cannot be guaranteed by the selection of tag of image based on location.Thus,this paper proposed the coverless information hiding based on partial-duplicate image retrieval.The specific research results are as follows:(1)A coverless information hiding method based on tags of position and partial-duplicate image retrieval.The concept of Non-Identical Near-Duplicate Image is utilized to communicate secret image by a series of partial-duplicate image.The database for coverless information hiding based on image is established by the construction of index based on partial-duplicate image and localized HSV feature of image,no longer the mapping relationship between image features and the secret message.Experimental results show that the proposed method can improve the hiding capacity of coverless information hiding based on image,and express the secret image by recovering the partial-duplicate image.However,the visual effects is limited because of the scale of our image dataset.(2)A coverless information hiding method based on chaotic encryption and partial-duplicate image retrieval.In order to improve the security of communicating the secret image,we encrypt the location tags of image by a key generation algorithm based on the logistic regression for that the proposed method can acquire the property of randomness,certainty,ergodicity,and sensitivity to initial conditions.Experimental results show that the proposed method can improve the security of the communication of secret image,can simplify the transmission of location tags,can improve the hiding capacity as well.