EPSMR: An efficient privacy-preserving semantic-aware multi-keyword ranked search scheme in cloud

Traditional privacy-preserving search schemes have commonly used the TF-IDF model, which relies on keyword frequency statistics but neglects the semantic relationships between and documents. In recent years, a few secure multi-keyword ranked search schemes over encrypted cloud data have considered the data semantics. However, the search efficiency of these schemes can still be improved. This paper presents an efficient privacy-preserving semantic-aware multi-keyword ranked search scheme (EPSMR) over encrypted cloud data. EPSMR utilizes the LDA topic model to discover hidden semantic information in documents and convert documents into vector representations. vector encryption by using the secure inner product is employed to implement the privacy-preserving semantic relevance score computation between search keywords and documents. To achieve efficient multi-keyword searches, the bisecting k-means algorithm and search tree structure are introduced to design a novel ω-way balanced search tree index which benefits finding search results quickly. to further improve the search efficiency, a filtering-threshold lookup table is presented, which can be used to prune unqualified sub-trees in advance. The experimental results show the superior performance of the proposed scheme compared to existing schemes in terms of both semantic search precision and search time cost.