An Efficient Distributed Approach on High Dimensional Data Similarity Searchable Encryption

Due to the rapid development of cloud computing, how to perform efficient similarity retrieval on high-dimensional encrypted data has extensive application value. And because the data size that many applications are facing has grown at a geometric level, traditional centralized algorithms can not meet the needs of users. In this paper, we present a distributed similar searchable encryption scheme for high-dimensional data. Our method takes advantage of local sensitive hashing (LSH) to reduce the dimensions and make a fast similar neighbor searching. Then random vector dot product and homomorphic encryption techniques are used to design a specific encryption index structures so that a more rigorous privacy definition can be satisfied. The index building and searching algorithms are all implemented under the MapReduce framework to ensure the high scalability during processing of massive data. Through the analysis of experimental results of real data, the presented scheme can effectively perform similar retrieval on cipher text while protecting privacy.