CRYPHTOR: A Memory-Unified NTT-Based Hardware Accelerator for Post-Quantum CRYSTALS Algorithms

Abstract

This paper presents the design and FPGA implementation of a hardware accelerator for the Post-Quantum CRYSTALS-Kyber and CRYSTALS-Dilithium algorithms, named CRYPHTOR (CRYstals Polynomial HW acceleraTOR). The proposed architecture includes a unified memory arrangement and dedicated ALUs for Kyber and Dilithium, capable of accelerating several polynomial operations such as Number Theoretic Transform (NTT), Inverse NTT, Coefficient-Wise Multiplication (CWM), modular addition and subtraction, modular reduction, and the multiply-accumulate operation. CRYPHTOR has been integrated into two SoCs: one based on a 64-bit RISC-V processor and the other on a 32-bit RISC-V microcontroller. In these configurations, up to 26x and 300x of speedup has been obtained for the NTT, and up to 30x and 140x of speedup for the matrix-vector multiplication compared to the software implementation running on the RISC-V processors.

Keywords

• Module learning with errors
• CRYSTALS-kyber
• CRYSTALS-dilithium
• FPGA
• postquantum cryptography