Design of Polynomial NTT and INTT Accelerator for Post-Quantum Cryptography CRYSTALS-Kyber

NIST post-quantum cryptography standardization round 3 announced CRYSTALS-Kyber as one of the finalists. As a lattice-based cryptography scheme, CRYSTALS-Kyber performance relies on polynomial multiplication efficiency. This paper presents a high-speed and pipelined hardware number theoretic transform (NTT) and INTT accelerator for CRYSTALS-Kyber. Our work centers around designing and optimizing the NTT accelerator architecture with suitable parameter for hardware implementations of CRYSTALS-Kyber. The work includes modifying the modular arithmetic modules and butterfly units structure with an efficient low-complexity algorithm. As a result, our design achieved 237 MHz f max when synthesized on Intel FPGA Cyclone V with Quartus. Resources utilization through combinational logic path rebalances allowed us to efficiently pipeline between hardware modules.