An Efficient NTT-Based Polynomial Multiplication Architecture for BFV Homomorphic Encryption

Homomorphic encryption (HE) ensures data security and privacy in cloud computing by permitting operations on encrypted information without revealing the original data. Reliable high-speed polynomial multiplication is an important operation to build HE and other lattice-based cryptography. There are two approaches in realizing polynomial multiplication:first school book multiplication and NTT-based polynomial multiplication. This paper proposed efficient NTT-based polynomial multiplication using improved folding Barrett modular multiplication (BMM) for Brakerski/Fan-Vercauteren (BFV) HE. We also implemented three types of BMM and integrated them into the polynomial multiplication to see and compare the performance. The result shows our architecture can reach 1.6× faster TPS compared to previous works.