Blockchain-based Privacy Preserving Linear Regression

Abstract

In this study we propose a blockchain-based architecture that uses smart contracts and homomorphic encryption to allow statistical computations on confidential data by third parties. The use of blockchain provides the much-desiredsecurity properties of integrity and fault tolerance and homomorphic encryption preserves the privacy of the data. We present the design, implementation, and testing of our system. Our results show that a blockchain-based data sharing mechanism with homomorphic calculations via a smart contract is feasible and provides improvements in protecting the data from unauthorized users. Even though our work focused on linear regression, the architecture can be used for other statistical analysis and machine learning algorithms.