An Efficient and Configurable Hardware Architecture of Polynomial Modular Operation for CRYSTALS-Kyber and Dilithium

The globalized migration towards post-quantum cryptography (PQC) is accelerating to protect communications against the upcoming quantum threat. However, the resource-constrained IoT communication devices limits the deployment multiple PQC algorithms, which is hard to fulfill the various security requirements of IoT devices. In this paper, an efficient hardware architecture compatible with customized instruction format is proposed, which is compatible with CRYSTALS-Kyber and Dilithium. Following the methodology of maximize-reuse, a configurable polynomial modular arithmetic unit is presented to execute all the required modes of polynomial modular operations in pipeline. Implemented on UltraScale+ and Artix-7 platforms, the proposed architecture consumes 2310 and 4028 equivalent slices at a maximum frequency of 280MHz and 99MHz in two platforms. Compared to the state-of-the-art Kyber-only and Dilithium-only researches, this work supports two algorithms simultaneously and realizes the lowest area-time (AT) value for Dilithium and a competitive AT value for Kyber.