Privacy preserving data analytics in financial inclusion and crowd computing

Abstract

Banks serve unbanked customers through financial inclusion, and other organizations use crowd computing to support their businesses. Personal and business data are being collected and used for such purposes, and data are usually processed in the clear on servers. Data owners do not have the means to protect their data from potential exposure beyond the intended purpose. This research aims to introduce new architectures and protocols for privacy-preserving data analytics which use Homomorphic encryption for data protection and computation. The first contribution optimizes financial inclusion by conducting a secured credit assessment with encrypted data on edge servers. It achieves reasonable prediction accuracy and response time. The second contribution adopts distributed crowd computing to train a logistic regression model with encrypted data and record activities on a blockchain. The generated prediction results are comparable to the predictive model trained using the full data set.