On GPU Implementations of Encryption Algorithms

Digital Object Identifier 10.32754/JMT.2019.2.04 29 1Abstract—With current evolution in technology and digitization process we see more often that data breaches are a real problem. There is a gap in data privacy because by using digital information processing we have three big zones of interest: data at rest, data in use and data in transit. The objective of this paper is to emphasize the importance of privacy and how data at rest and data in transit can be protected using cryptographic primitives. By using graphical processing units (GPUs) and special processor instructions (AES-NI) we accelerated some of the mainly used symmetric encryption schemes. By comparing the results with classical CPU implementations, we obtained for AES-128 a speedup of approximately x346 in OpenCL implementation and x123 by using CUDA framework. When having compared the same result with the AES-NI instruction set version, we got an improvement factor of 1.6. We also accelerated 3DES using the same frameworks and we managed to quicken the encryption with a speedup of x1101 using CUDA, respectively x2365 for OpenCL.