High-Performance NTRU Accelerator Using a Direct Memory Access Controller

Abstract

NTRU is a well-established and widely used public-key cryptography. It has been standardized in IEEE Std1363.1 and X9.98, and its variant was submitted to the National Institute of Standards and Technology (NIST) post-quantum cryptography (PQC) standardization contest. In this study, we proposed a high-performance accelerator implementation of the recent NTRU version submitted to the NIST PQC contest. In NTRU, there are two types of polynomials: ternary polynomial (TP) that has coefficients in {-1, 0, 1}, and non-TP. In contrast to previous methods that support both multiplication of a TP and a non-TP and multiplication of non-TPs, the proposed accelerator performs only multiplication of non-TPs by converting a TP to a non-TP in advance. We also leveraged a direct memory access controller and simplified the definitions of the polynomial registers. Consequently, we reduced the latency, data transfer time, and resources, such as the need for a large number of registers and multiplexers. We synthesized the accelerator using 28-nm CMOS process technology. The results showed that the accelerator can perform encryption and decryption within 0.8 and 1.8 mu s, respectively, at the maximum clock frequency of 1.67 GHz, requiring 494.9 k gate counts. Furthermore, the proposed accelerator demonstrated a higher performance with fewer resources than the existing accelerators.