Abstract:Since traditional cryptosystems have been vigorously challenged in recent years by quantum computing, using multivariate systems to design cryptosystems becomes a possible choice among those post-quantum candidates. However, compared with traditional cryptosystems (RSA, ECC, etc.), multivariate systems might not be cost-friendly for practical applications. In recent years, GPU is widely used in machine learning with respect to its massive parallel computing power. To the best of our knowledge, all the published GPU acceleration schemes for multivariate systems are based on quadratic multivariate systems, and they might not be applicable for high-order multivariate cryptography systems. In this paper, we propose a generic GPU acceleration framework for multivariate systems with various orders. The experiment results show that our optimization method can effectively improve the performance of multivariate cryptosystems on GPUs.

A Generic Optimization Method of Multivariate Systems on Graphic Processing Units