High-efficiency low-latency ntt polynomial multiplier for ring-lwe cryptography

This paper presents a novel architecture to perform polynomial multiplication in ring learning with errors (ring-LWE) cryptosystems. By employing number theoretic transform (NTT) of the input polynomials simultaneously, the multiplication latency is significantly reduced. In addition, a multiple-path delay feedback (MDF) architecture is used to speed up the multiplication process. As a result, the proposed NTT multiplier offers a better value of area-latency product compared with that of previous studies. The simulation results for the security parameters n = 512 and q = 12,289 on Xilinx Virtex-7 FPGA show that the proposed multiplier uses only about 8.69% of the number of clock cycles required by previous works to completely perform the polynomial multiplication. Furthermore, the obtained area-latency product value of the proposed architecture is less than 45.3% of that of previous works.