Research Of Privacy-Preserving Multi-Keyword Search Schema Over Encrypted Cloud Data

Cloud data owners are usually inclined to encrypt data before outsourcing it to service providers in order to achieve the purpose of protecting privacy.However,most of the existing indexing schemes improve retrieval efficiency at the cost of reducing the accuracy.How to design a ciphertext indexing scheme which can not only improve the retrieval efficiency but also ensure the accuracy of the retrieval results is particularly significant.MRSE-BBKS scheme and its extension scheme MRSE-EBKS are proposed in this paper.The multi-tree index mechanism MRSE-BBKS scheme based on dichotomous k-means is proposed by combining the Bisecting k-means method and the multi-tree method,which constructs the index tree from the bottom to the top.At the same time,Greedy Depth-first Search algorithm is proposed by calculating the correlation score between pruning vectors and query vectors in order to prun the unrelated document clusters to a great extent,which not only ensures the accuracy of retrieval results,but also greatly improves the efficiency of retrieval and saves the cost of index space.The MRSEEBKS scheme uses the index construction method in MRSE-BBKS,while supporting multi data owners' outsourcing service instead of supporting single data owner's outsourcing service.The MRSE-EBKS scheme also puts forward Document Authority Model to weigh index vectors,and introduces Relevance Feedback Model to optimize retrieval vectors as well,so as to achieve secondary retrieval,so that authorized users can obtain more authoritative,self-satisfied and most relevant high-quality retrieval results.Theoretical analysis and experimental results show that MRSE-BBKS scheme outperforms the related ciphertext multi-keyword retrieval method,and the cost of index space is obviously saved,which can achieve safe and highly efficient multi-keyword ciphertext retrieval.MRSE-EBKS retrieval scheme can return more authoritative high-quality related documents,which helps to improve document recalling rate and realize personalized retrieval.