Design And Implementation Of Parallel Encrypted DBMS Based On CryptDB

With the rapid development of Internet technology,more and more users interact with the servers every day and everywhere.Those interactions generate a huge amount of data on the Internet.In addition,the popularity of personal SkyDrive makes users tend to save data in the cloud disk that cost cheaper.These data maintain users' ID card information,cell phone numbers,credit card numbers,medical information and other privacy data.All the companies who provide services have an obligation to protect the users' privacy information when they collect those data.Therefore,it has become a hot research topic to design and implement an encrypted DBMS with high performance and can provide the service to search data without encrypting them.This is a meaningful research.According to process of software design and development,this thesis use use-case diagram to do system requirement analysis and abstract the system's function modules.The activity diagram is used to model the behavior of the system,and determine the whole process of the system.Then,the interaction and dependence of each module are determined by the sequence diagram.Finally,we use java programming language combined with MySQL-Proxy to achieve a simple ciphertext database system based on Crypt DB's onion encryption model.In order to improve the usability of the system,we design and implement a friendly man-machine interactive website based on the lightweight web application framework named Flask.Combining the characteristic of MapReduce and the encrypted DBMS which based on Crypt DB system,a parallel insert and distributed storage schema is proposed to improve the performance of encrypting,and another two algorithms named ‘JobControl' and ‘FileSplit' are also proposed to improve the parallelism and controllability of this schema.In this thesis,the ‘map' method of MapReduce is rewritten and the performance of the schema is analyzed at the end.The experimental results show that the speed-up radio of this proposed scheme can reach 2.51,and the total time cost of encrypted DBMS parallel schema can be reduced to 39.8% on the cluster consisting of 3 encrypted DBMS nodes.The analysis and experimental results show that the schema has a high degree of paralleling,and it can be used in the encryption and storage of large-scale relational data.