Implementation of Secure, Fast and Efficient RSA Algorithm on GPU

Abstract

In the field of cryptography, public key algorithms satisfy all the four requirements of security: Confidentiality, Integrity, Non-repudiation, Authentication as compared to the conventional cryptosystem. Public key cryptography is widely known to be slower than symmetric key alternatives for massive computations of modular arithmetic. The modular arithmetic and modular exponentiation makes RSA computationally expensive when compared to symmetric algorithms. Therefore, to make a more efficient, secure and faster implementation of RSA algorithms is publicly concerned. With the development of the GPGPU (General-purpose computing on graphics processing units) field, more and more computing problems are solved by using the parallel property of GPU (Graphics Processing Unit). NVIDIA’s proprietary CUDA and the Khronos Group’s open standard OpenCL are the two frameworks designed to exploit the massive parallelism . The target in this work is to develop a secure, faster and efficient RSA using CUDA and OpenCL. Finally, this work will compare the performance between the CUDA-RSA and CPU-RSA, OpenCL-RSA and CPU-RSA and CUDA-RSA and OpenCL-RSA for small and large prime numbers, in order to look into the possibility of improving the performance of RSA algorithms