A Design Of Image-Oriented Cloud Storage Data Deduplication

In current Internet age,the rapid development of cloud storage services makes the explosive growth of clients' data.Client-side deduplication technology that can save storage space and keep networkbandwidth down is widely used.It stores only a single copy of repeating data on the server.Due to the characteristics of large quantity and redundancy,it is especially necessary to perform client-based deduplication of images.However,the existing clientbased deduplication schemes are mostly based on cryptographic hash,which only support exact data deduplication,and are not suitable for images deduplication.For images,typical modifications such as resize,compression,and format conversion only change their binary presentation but maintain the perceptual content,which shoule be treated duplicate data.Therefore,it needs to consider image fuzzy deduplication for perceptual similar images.On the other hand,in order to achieve secure client-side deduplication,the Proofs of Ownership(Po W)that can help the cloud server to verify whether a client indeed has a complete original file works well in file deduplication.But the previous Po W schemes cannot be applied to the image deduplication any more.In order to implement an efficient and secure client-based image deduplication,the existing image deduplication schemes are studied and analyzed in this paper,and the main research results obtained are as follows:1.A client-based image deduplication scheme supporting fuzzy matching is proposed.Firstly,theoretical analysis and experimental verification of discrete cosine transform based image perceptual hash algorithm are showned in this paper,and according to the statistical properties of discrete cosine transform spectral coefficients,an improved perceptual hash algorithm is designed based on the advantage of the DAN perceptual hash(DAN-phash)by utilizing Zigzag coding.Secondly,it gives the definition of the perceptual similar images in this scheme and clarifies the goal of image deduplication.Thirdly,the client-based image deduplication scheme matches similar images and achieves the purpose of image deduplicationby applying the improved algorithm.Finally,the performance tests and comparison with existing schemes show that the algorithm's false deduplication rate for different images is only 0.046% and recognition rate of similar images can reach 88.89% when the threshold value is 0.017.In practical applications it can achieve a lower false deduplication rate for content modified images by setting suitable thresholds under the slightly higher deduplication ratio,also improve the matching accuracy,reliability and the anti attack capability.2.A pseudo-random block based proof of ownership for similar images scheme is proposed.It completes the Po W process according to the challenge information of the cloud server and the client's proof information based on the original image.In order to improve security,it introduces pseudo-randomness in the Po W process to guarantee the freshness of proof information,protecting the scheme from attacks like the replay attack of proof and resolving the problem that attacker fools the server to obtain the image illegally only by the perceptual hash.In the premise of ensuring the security,it does not need to rely on the auxiliary image libraries,effectively reducing the storage burden of the server and the client.