Cryptographic Primitives Optimization Based on the Concepts of the Residue Number System and Finite Ring Neural Network.

Data encryption has become a vital mechanism for data protection. One of the main challenges and an important target for optimization is the encryption/decryption speed. In this paper, we propose techniques for speeding up the software performance of several important cryptographic primitives based on the Residue Number System (RNS) and Finite Ring Neural Network (FRNN). RNS&FRNN reduces the computational complexity of operations with arbitrary-length integers such as addition, subtraction, multiplication, division by constant, Euclid division, and sign detection. To validate practical significance, we compare LLVM library implementations with state-of-the-art, high-performance, portable C++ NTL library implementations. The experimental analysis shows the superiority of the proposed optimization approach compared to the available approaches. For the NIST FIPS 186-5 digital signature algorithm, the proposed solution is 85% faster, even though the sign detection has low efficiency.