Accelerating Homomorphic Encryption using Systolic Arrays with Polynomial Optimization

Internet traffics are growing more dense day by day. Internet traffics also represent the privacy of many people. Other than privately-owned data, IoT devices and automation also send their perceived data in its unencrypted form. Cryptography can deliver integrity and data security that are much needed to secure the private data of many people in this world. However, a problem still exists in using cryptography algorithm. As more secure data is, it is far harder for its algorithm to be decoded in linear time complexity. In this paper, we develop an architecture for accelerating BFV homomorphic encryption that can handle data operation without being decrypted. By using systolic arrays and polynomial optimization, we accelerate the time executions for polynomial degrees 4, 8 and 16. Various experiments on Intel I5, RasPi 3B+, and PYNQ FPGA are also shown. The design is useful for secure communication and computing.