Journal
INFORMATION SCIENCES
Volume 575, Issue -, Pages 217-230Publisher
ELSEVIER SCIENCE INC
DOI: 10.1016/j.ins.2021.06.017
Keywords
Privacy-preserving computing framework; Cloud computing; Homomorphic encryption; Multiple keys
Categories
Funding
- Research Start-up Projects for High-level Talents [000527]
- Research Start-up Projects of Shenzhen University [000002110102]
- Basic Research Project of Shenzhen, China [JCYJ20180507183624136]
- National Natural Science Foundation of China [61872109]
- Shenzhen Basic Research Project (Key Program) [JCYJ20200109113405927]
Ask authors/readers for more resources
Preparing large amounts of training data is crucial for the success of machine learning, while privacy-preserving techniques like homomorphic encryption are proposed to address individual privacy concerns. Collaboration between different institutions is common in the era of big data, but there are risks to data privacy when encrypting data under a single key in multi-institution scenarios.
Preparing large amounts of training data is the key to the success of machine learning. Due to the public's concern about individual privacy, different techniques are proposed to achieve privacy preserving machine learning. Homomorphic encryption enables calculation on encrypted data in the cloud. However, current schemes either focus on single key or a specific algorithm. Cooperation between different institutions is quite common in this era of big data. Encrypting data from different institutions under one single key is a risk to data privacy. Moreover, constructing secure scheme for a specific machine learning algorithm lacks universality. Based on an additively homomorphic encryption supporting one multiplication, we propose a general multikey computing framework to execute common arithmetic operations on encrypted data such as addition, multiplication, comparison, sorting, division and etc. Our scheme can be used to run different machine learning algorithms. Our scheme is proven to be secure against semi-honest attackers and the experimental evaluations demonstrate the practicality of our computing framework. (c) 2021 Elsevier Inc. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available