Optimized ciphertext retrieval for cloud computing based on dynamic clustering

With the extensive application of cloud storage, the amount of data stored in the server rapidly increases. At the same time, user documentation typically contains privacy-sensitive information that needs to be encrypted before uploaded to the cloud server. Facing such a large number of ciphertext data, the existing search technologies in retrieve of large amounts of data efficiency of the ciphertext data are fulfilling. To solve this problem, this paper makes improvement based on similarity search tree ciphertext retrieval method (MRSE-SS). In the big data environment, this paper proposes dynamic range clustering algorithm DIK-MEDOIDS. This method divides gap between the maximum and minimum document vectors into k slots. The size of the slots is hypersphere diameter, and the document vector which is closest to the middle value of the range is set as the hypersphere center. The size of each documentation slot depends on number of documents. Different with MRSS-SS algorithm, the proposed algorithm can search the qualified documents in the searching stage among hypersphere, to expand the cover of both documents and searching sets for returned data more in line with the user search need. With the increase of the documents, slots are dynamically determined. At the search stage, the search vector is represented as a hypersphere. Cloud server makes the determination through the position of search vector hypersphere and similarity search tree nodes. Along with the number of documents increasing, initialized time increases linearly, while the search stage is same with algorithms MRSE-SS which is shown in formula. When the document vector DC exponential grows, the time complexity is O (l).