GPU-acceleration of blowfish cryptographic algorithm

Abstract

The demand for fast and secure cryptographic encryption techniques has been growing over the recent years. The introduction of the CUDA programming framework by NVIDIA allowed utilizing GPUs for general purpose computing. The use of GPUs in the cryptography domain has become quite popular ever since. In this paper, we present a CUDA implementation of the blowfish encryption algorithm. The algorithm has been designed to make use of the unified memory model introduced as a part of CUDA 6. Experimental results show that the unified implementation of the blowfish algorithm performs better than an efficient CPU implementation by a factor of 62X and performs twice better than a non-unified CUDA implementation of the algorithm. We also present possible solutions to achieve an optimal configuration for the target GPU to get the best performance.