A Ranked Searchable Encryption Scheme Based On MapReduce

With the rapid development of cloud computing, more and more users choose to storetheir data in cloud server. At the same time, they begin to focus on user privacy and datasecurity. Considering data security and user privacy, it is necessary for users to encrypt theirsensitive data before moving them to the cloud server. However, it becomes inconvenientwhen the users attempt to retrieve the documents in the cloud. Searchable encryption schemesprovide an important mechanism to cryptographically protect data and make it available to besearched and accessed. Although most searchable encryption schemes allow users to securelysearch over encrypted data through keywords, these schemes will return all matcheddocuments without filtering them. What’s more, few ranked searchable encryption schemesconsidered users’ feedback problems after download and the difference of keywords at thebeginning. And the computational overhead of calculating keyword’s weight and buildingindex is done by data owners. It makes searchable encryption fail to be efficient and accurate.As a statistical method, TF/IDF is used in information retrieval and text mining area. Itcan evaluate the importance of a set of keywords to a file in the corpus. As a simple way tomeasure the correlation the user’s queries in the corpus, its various forms have often been usedin the search engine. If a keyword has a high frequency in a file and less frequency in otherfiles, the keyword will be used to distinguish.As a programming model for processing and generating large data sets with a paralleland distributed algorithm, MapReduce is usually used in searching area. It can resolve thecalculation problem of huge amounts of data. A MapReduce program is composed of a Mapand a Reduce. Map procedure performs filtering and sorting and Reduce procedure performs asummary operation. In order to make an implement of ranked searchable encryption schemeefficiently and safely, we choose MapReduce.The main contributions of the dissertation are listed as follows:a. This paper proposes a ranked searchable encryption scheme based on MapReduce. Thevalue of keyword is calculated by data owner. It will return to users the top-k documentsfrom server ordered by this value.b. This paper proposes an improved scheme. Considering the initial weights of differentkeywords, we will modify the weight at the beginning. In order to reduce the cost of local computation, we will move a part of calculation of the weight to cloud server. With theconsideration of the safety factors, we choose relative frequency instead of the truefrequency. For making users more satisfied, we add feedback mechanism.c. At the end of this article, it makes an implement of these two schemes by MapReduce.After comparison and discussion between these two schemes, we can see improvedscheme is more suitable for the requirements of users.