Construction of a Fully Homomorphic Encryption Scheme with Shorter Ciphertext and Its Implementation on the CUDA Platform.

Dong Chen,Tanping Zhou,Wenchao Liu,Zichen Zhou,Yujie Ding,Xiaoyuan Yang
DOI: https://doi.org/10.1007/978-3-031-26281-4_14
2023-01-01
Abstract:Homomorphic encryption supports meaningful operations on ciphertext and is widely used in outsourcing computing, secure multi-party computing and other scenarios. The TFHE-type fully homomorphic encryption scheme has an efficient bootstrapping process, so it supports efficient homomorphic logic gate circuits. However, its single-bit encryption feature results in a large ciphertext expansion (the ciphertext expansion rate of the CGGI16, CGGI17, and ZYL + 18 scheme is 16032). Therefore, how to reduce the scale of ciphertext without affecting the efficiency is a practical problem. TFHE-type fully homomorphic encryption scheme with shorter ciphertexts is constructed. By applying the round function to the encryption process, the ciphertext scale is reduced by 62%; Using the GPU's ability to implement large-scale matrix operations efficiently, a fully homomorphic encryption scheme with shorter ciphertexts is implemented based on the cuFHE. The experimental results show that, compared with the CGGI17 scheme, the running time of a single gate circuit (including the bootstrapping process) in this scheme does not exceed 1 ms on the CUDA platform, and the ciphertext scale of this algorithm is reduced by 62%.
What problem does this paper attempt to address?