Privacy-preserving distributed analytics: Addressing the privacy-utility tradeoff using homomorphic encryption for peer-To-peer analytics

Abstract

Data is becoming increasingly valuable, but concerns over its security and privacy have limited its utility in analytics. Researchers and practitioners are constantly facing a privacy-utility tradeoff where addressing the former is often at the cost of the data utility and accuracy. In this paper, we draw upon mathematical properties of partially homomorphic encryption, a form of asymmetric key encryption scheme, to transform raw data from multiple sources into secure, yet structure-preserving encrypted data for use in statistical models, without loss of accuracy. We contribute to the literature by: I) proposing a method for secure and privacy-preserving analytics and illustrating its utility by implementing a secure and privacy-preserving version of Maximum Likelihood Estimator, "s-MLE", and ii) developing a web-based framework for privacy-preserving peer-To-peer analytics with distributed datasets. Our study has widespread applications in sundry industries including healthcare, finance, e-commerce etc., and has multi-faceted implications for academics, businesses, and governments.