Accelerating Falcon on ARMv8

Falcon is one of the promising digital-signature algorithms in NIST’s ongoing Post-Quantum Cryptography (PQC) standardization finalist. Computational efficiency regarding software and hardware also main criteria for PQC standardization. In this paper, we present an efficient implementation on ARMv8 environment. Until now, most optimization have been conducted 32-bit ARM (Cortex-M4) typical CPUs (Intel AMD CPUs). However, including Cortex-A30, 50, 70 series widely used various IoT (Internet Things) applications, Edge computing devices, OBUs (On Board Units) autonomous driving cars. For optimizing performance Falcon, take full advantage NEON engine which a kind parallel processing unit MCU. The computation belongs to polynomial multiplications Complex number domain Integer domain. Typically, FFT (Fast Fourier Transformation)-based multiplication method NTT (Number Theoriteic Transform)-based domain, respectively. Thus, order enhance overall improve FFT-based NTT-based by utilizing ARMv8. Specifically, parallelize process (FFT/NTT transformation, pointwise multiplication, inverse FFT/NTT transformation) with strategically vector instructions. Furthermore, minimize redundant memory accesses during FFT/NTT-based making available registers engine. Through proposed multiplications, provides 15.1% (resp. 18.1%), 16.5% 17.1%), 65.4% 69.4%) improvement keypair generation, signing, verification at security level 1 5) compared reference submitted final round NIST competition. as far know, first optimized