Privacy Preservation in Multi-Cloud Secure Data Fusion for Infectious-Disease Analysis

Abstract

It is often observed that people's data are scattered across various organizations and these data can be used to generate usable insights when integrated. However, data fusion from multiple data hosting sites could put user privacy at risk albeit with some security mechanisms. This paper studies a data-analytic platform that adopts the Kulldorff scan statistic to determine infectious-disease spatial hotspots by integrating and analyzing users' health and location data that are respectively stored in two clouds. We examine the privacy threats to this platform which has a key-oblivious inner product encryption (KOIPE) mechanism in place to ensure that only coarse-grained statistical data is revealed to the honest-but-curious (HbC) entity. To protect user privacy from the designed inference attack, we exploit a game-theoretic approach to incentivize users to form anonymous clusters with a quantitative privacy guarantee. We conduct extensive simulations based on real-life datasets to demonstrate the performance of our scheme in terms of design overhead and privacy level.