Secure Similarity Search Over Encrypted High-Dimension Data

As cloud computing develops,the application of cloud storage is becoming more and more common.However,cloud service providers are semi-trusted.Therefore,while enjoying the convenience brought by cloud storage,ensuring data security on cloud storage is an important part that cannot be ignored.In order to achieve this goal,searchable encryption technology emerges as the times require.It can not only ensure the data security,but also realize the search function.Further optimizing the search performance and improving the security are the goals of the searchable encryption field.This paper focuses on the similarity search of high-dimension data in searchable encryption,and finds the following problems from two aspects: search and security.In terms of search,it is found that when the data is not evenly distributed,the structure of hash table as index will be unbalanced,which will lead to the decline of query quality.In terms of security,at present,the highest practical security level is forward security,while most of the current schemes that can achieve forward security are for keywords,and very few for high-dimension data.To solve the above problems,this paper proposes two corresponding schemes.Both schemes use searchable encryption and locality-sensitive hashing algorithm.In the scheme of optimizing search performance,the index based on data distribution is constructed,and the appropriate index location is selected for each data point according to data distribution density,so as to improve the imbalance of original index structure and query quality.In the scheme of improving the security level,by separating the hash table and the list in the traditional index structure,and connecting all the lists with their stored encrypted information,an index that can provide forward security is built.In this paper,detailed security proofs and sufficient experimental analysis of the two schemes are carried out,and the results show that the two schemes can effectively optimize the search performance and improve the security.