Achieving Efficient Similar Document Search over Encrypted Data on the Cloud

Abstract

Cloud computing, which provides scalable computing resources at an economical rate, is more important than ever. More and more data owners are storing their data on the cloud. To protect private and sensitive data, user data should be encrypted by the data owner before it is outsourced to the cloud. However, this defeats the merits of cloud computing; the data needs to be decrypted and consumed on the client side. In this paper, we introduce a practical searchable encryption scheme which supports keyword search and similar document search, based on the Vector Space Model (VSM), by harnessing the power of homomorphic encryption (HE). HE is an encryption scheme where arithmetic calculations can be performed without decryption. We first build a term index tree to filter out irrelevant documents. Subsequently, we perform cosine similarity calculation upon search requests. Additionaly, we propose a parallel version of this scheme, which can effectively utilize the power of the cloud. Experiments on real-world datasets indicate that our scheme can effectively provide practical keyword search based on VSMs in a cloud environment.