December 1, 2022
Duc Tri Nguyen - George Mason University
We present our speed records for Falcon signature generation and verification on ARMv8-A architecture. Our implementations are benchmarked on Apple M1 ’Firestorm’ and Raspberry Pi 4 Cortex-A72 chips. Compared with lattice-based CRYSTALS-Dilithium, our optimized signature generation is 2× slower, but signature verification is 3–3.9× faster than the state-of-the-art CRYSTALS-Dilithium implementation on the same platforms. Compared with stateful and stateless hash-based digital signatures, XMSS and SPHINCS+, our optimized Falcon implementation has 10–950× faster signature generation and 23–31× faster signature verification. Faster signature verification may be particularly useful for the client side on constrained devices. Our Falcon implementation outperforms the previous work in both Sign and Verify operations. We achieve improvement in Falcon signature generation by supporting all possible parameters N of FFT-related functions and applying our compressed twiddle-factor table to reduce memory usage. We also demonstrate that the recently proposed signature scheme Hawk, sharing functionality with Falcon, offers 17% smaller signature sizes, 3.3× faster signature generation, and 1.6–1.9× slower signature verification when implemented on the same ARMv8 processors as Falcon.
