Parallel Generation of RSA Keys

Abstract

With massive growth in technology, online communication among different entities now encompasses the transfer of confidential data. It is due to this confidentiality and data security that the need of encryption algorithm arises. Being one of the most widely used asymmetric or public key cryptographic algorithms, the RSA algorithm does require substantial amount of study, improvement and management so as to get the best possible efficiency in terms of execution time and memory utilization. Key Generation being the first stage of RSA requires to be the most time efficient step for commercial deployment where the requirement of keys is very high. The size of the key cannot be compromised for faster execution as it can prove to be an hindrance to the security level of RSA. Hence, this work aims towards the parallel implementation of the key generation process of RSA by using the parallel programming paradigm of CUDA. A comparative analysis of the sequential RSA implemented in C and its modified parallel approach implemented in CUDA has been shown graphically. The same implementation has also been done by using multiple precision libraries to support key generation of larger key sizes. Results of implementation using the GMP and MP libraries for various key sizes have also been generated.