Research Article
Privacy-Preserving Computation Tookit on Floating-Point Numbers
@INPROCEEDINGS{10.1007/978-3-030-89814-4_33, author={Zekai Chen and Zhiwei Zheng and Ximeng Liu and Wenzhong Guo}, title={Privacy-Preserving Computation Tookit on Floating-Point Numbers}, proceedings={Mobile Multimedia Communications. 14th EAI International Conference, Mobimedia 2021, Virtual Event, July 23-25, 2021, Proceedings}, proceedings_a={MOBIMEDIA}, year={2021}, month={11}, keywords={Privacy computation Homomorphic encryption Multiple keys Secure computation}, doi={10.1007/978-3-030-89814-4_33} }- Zekai Chen
Zhiwei Zheng
Ximeng Liu
Wenzhong Guo
Year: 2021
Privacy-Preserving Computation Tookit on Floating-Point Numbers
MOBIMEDIA
Springer
DOI: 10.1007/978-3-030-89814-4_33
Abstract
Computation outsourcing using virtual environment is getting more and more prevalent in cloud computing, which several parties want to run a joint application and preserves the privacy of input data in secure computation protocols. However, it is still a challenging task to improve the efficiency and speed of secure floating point calculations in computation outsourcing, which has efficient secure integer calculations. Therefore, in this paper, we propose a framework built-up with a privacy-preserving computation toolkit with floating-point numbers (FPN), called PCTF. To achieve the above goal, we provide efficient toolkit to ensure their own data that FPN operations can be securely handled by homomorphic encryption algorithm. Moreover, we provide simulation results to experimentally evaluate the performance of the accuracy and the efficiency of PCTF, which will slowdown with 10x time consumption per in the secure floating-point addition and secure floating-point multiplication. Existing FPN division is constantly approaching result of division, or obtaining the quotient and remainder of division, in terms of precision, it is impossible to guarantee the precise range stably. However, our PCTF has higher precision in secure floating-point division, and the precision can be guaranteed at least({10^{-17}}).
