An Efficient and Secure Database System using OpenCL

Abstract

Nowadays, outsourced database systems have been widely used by organizations due to its low cost and easy availability. But security is always a big concern for these databases. Data Encryption is one of the many ways to secure data in database systems. It encrypts data into cipher text and decrypts it back into plaintext with the help of cipher keys. But it is an expensive process and takes more time when applied on large amount of data. With evolution of computer hardware, General purpose computing on graphic processing units is used in many research areas due to its massive parallel computing power. OpenCL and CUDA frameworks are used for general purpose computing on GPUs. In this thesis-work, an efficient database system has been implemented which uses GPU for encryption and decryption of data fields. It is mainly designed to provide a high security to the database systems. Also, it has been proved that for small sized data, encryption on CPU is more efficient and fast as compared to encryption on GPU. Therefore, the main aim of this work is to achieve a high overall performance of a database system by including a mechanism which will decide whether to exploit the computational capability of GPU or not. Advanced Encryption Standard (AES) has been implemented using OpenCL on AMD GPUs and then integrated with the MySQL database on Windows. The implementation supports 128 bit, 192 bit and 256 bit keys. The evaluation shows that the gain in the performance is dependent on the size of data and our proposed system enhances overall performance of the database system by transferring the encryption and decryption of large sized data to GPU.