Analytics on Anonymity for Privacy Retention in Smart Health Data

Abstract

Advancements in smart technology, wearable and mobile devices, and Internet of Things, have made smart health an integral part of modern living to better individual healthcare and well-being. By enhancing self-monitoring, data collection and sharing among users and service providers, smart health can increase healthy lifestyles, timely treatments, and save lives. However, as health data become larger and more accessible to multiple parties, they become vulnerable to privacy attacks. One way to safeguard privacy is to increase users’ anonymity as anonymity increases indistinguishability making it harder for re-identification. Still the challenge is not only to preserve data privacy but also to ensure that the shared data are sufficiently informative to be useful. Our research studies health data analytics focusing on anonymity for privacy protection. This paper presents a multi-faceted analytical approach to (1) identifying attributes susceptible to information leakages by using entropy-based measure to analyze information loss, (2) anonymizing the data by generalization using attribute hierarchies, and (3) balancing between anonymity and informativeness by our anonymization technique that produces anonymized data satisfying a given anonymity requirement while optimizing data retention. Our anonymization technique is an automated Artificial Intelligent search based on two simple heuristics. The paper describes and illustrates the detailed approach and analytics including pre and post anonymization analytics. Experiments on published data are performed on the anonymization technique. Results, compared with other similar techniques, show that our anonymization technique gives the most effective data sharing solution, with respect to computational cost and balancing between anonymity and data retention.