科研成果详情

With the development of quantum computers, NIST started post-quantum cryptography standardization to design post-quantum-secure cryptographic algorithms. Saber is a cryptosystem in third-round public-key encryption and key-establishment algorithm finalists. Because of its power of 2 number theoretic transform (NTT)-unfriendly ring, originally, Karatsuba polynomial multiplication algorithm (KPMA) and Toom-Cook polynomial multiplication algorithm (TCPMA) are used to speed up its computation-intensive matrix-vector multiplications. In later studies, NTT-based methods are applied to Saber on ARM platforms and result in a 61% speed-up. This work aims at adapting Saber with existing polynomial multiplication algorithms (PMAs), including non-NTT-based and NTT-based PMAs, to two energy-efficient RISC-V development boards, SiFive HiFive1 Rev B as well as Terasic T-Core. A 32-bit multiplier adapting Barrett reduction is designed to solve the overflow problem caused by RISC-V platform limitation. Experiment results show that the computation complexity depends on the algorithm choice and the underlying platform. NTT-based algorithms analytically have obvious advantages compared with non-NTT-based PMAs. However, the on-board cycle count on T-Core shows that NTT-based algorithms may have no comparability with non-NTT-based algorithms due to the high complexity overflow solutions. In addition, using the newly designed 32-bit multiplier can result in a 36.4% speed-up in practice. These results suggest several criteria for selecting algorithms on different platforms. This project can serve as a reference for future exploratory studies.