Research On The Method Of Image Fuzzy Deduplication Supporting Proof Of Ownership

In the age of cloud storage,a large number of users store their data on the cloud server,which produces a lot of redundant data.An effective technology that removes redundant data is deduplication.It stores only a single copy of repeating data on the server,which can save bandwidth and space.Researchers have proposed a lot of client-based deduplication methods of text files.Because images have the characteristics of large quantity and redundancy,it is necessary to perform image deduplication.However,most of the deduplication methods of text files cannot apply to image deduplication.The client-based deduplication of images faces many new challenges.Firstly,image deduplication needs to support fuzzy duplicate check.However,most of the existing deduplication methods use hash values of files to check duplicate precisely.These methods cannot respond to the new challenge of checking duplicate fuzzily.Secondly,image deduplication needs to support proof of ownership for similar images.However,none of the existing literatures takes the ownership verification of similar images into consideration.Thirdly,image deduplication needs to assess and compare image quality.However,image quality assessment is not taken into consideration in all of the researches about client-based image deduplication.Aiming at these new challenges,a scheme named a client-based image fuzzy deduplication method supporting proof of ownership is proposed in this paper.The proposed scheme consists of three phases: duplicate check,proof of ownership and quality comparison.In the duplicate check phase,the deduplication system needs to check duplicate fuzzily for similar images.The discrete cosine transform based perceptual hash has a high discriminative capability to images,however,the accuracy of duplicate check using this algorithm is not enough.Therefore,a high accuracy perceptual hash algorithm based on the discrete cosine transform is designed to meet the high accuracy requirement of fuzzy duplicate check.In the proof of ownership phase,the system needs to perform ownership verification for similar images,and no existing research involves this issue.Therefore,a new protocol named the perceptual hash based proof of ownership protocol is firstly proposed to ensure that only the one who owns the image indeed can pass the verification.In the quality comparison phase,the system needs to evaluate and compare quality of the client's image to that of the image stored on the server.Because the deduplication is client-based,the scheme needs a no-reference quality assessment method.Therefore,a no-reference image quality assessment method supporting distortion identification is designed to complete the evaluation and comparison of image quality.Security analysis results demonstrate that security strength of the proposed scheme is provable.A large number of simulation results show that the scheme can check image similarity accurately and assess image quality of a variety of distortions,meeting these new technical challenges.Moreover,performance test results show that time cost of the scheme is small,accomplishing image deduplication quickly and efficiently,and saving a lot of bandwidth and space needed to upload and store duplicated data.