Timing-Optimized Hardware Implementation to Accelerate Polynomial Multiplication in the NTRU Algorithm

Abstract

Post-quantum cryptographic algorithms have emerged to secure communication channels between electronic devices faced with the advent of quantum computers. The performance of post-quantum cryptographic algorithms on embedded systems has to be evaluated to achieve a good trade-off between required resources (area) and timing. This work presents two optimized implementations to speed up the NTRUEncrypt algorithm on a system-on-chip. The strategy is based on accelerating the most time-consuming operation that is the truncated polynomial multiplication. Hardware dedicated modules for multiplication are designed by exploiting the presence of consecutive zeros in the coefficients of the blinding polynomial. The results are validated on a PYNQ-Z2 platform that includes a Zynq-7000 SoC from Xilinx and supports a Python-based programming environment. The optimized version that exploits the presence of double, triple, and quadruple consecutive zeros offers the best performance in timing, in addition to considerably reducing the possibility of an information leakage against an eventual attack on the device, making it practically negligible.