Authors: Douwei Lei, Debiao He, Cong Peng, Min Luo, Zhe Liu, Xinyi Huang

Title: Faster Implementation of Ideal Lattice-based Cryptography Using AVX512

Journal: ACM Transactions on Embedded Computing Systems

Abstract: With the development of quantum computing, the existing cryptography schemes based on classical cryptographic primitives will no longer be secure. Hence, cryptographers are designing post-quantum cryptographic (PQC) schemes, and ideal lattice-based cryptography has emerged as a prime candidate. Today, as ideal latticebased cryptography becomes more mature, its performance becomes an important optimization goal. In ideal lattice-based cryptography, polynomial arithmetic and polynomial sampling are the most time-consuming operations and therefore need to be accelerated. In this paper, taking advantage of the parallelism of new 512-bit advanced vector instructions (AVX512), we present parallel implementations of polynomial arithmetic and polynomial sampling, thus comprehensively improving their performance. We conduct experiments with the Dilithium scheme(one scheme of NIST PQC Standardization Process Round-4). Our implementation gets a nice performance boost compared to its pure C language and 256-bit advanced vector instructions (AVX2) implementation.