Survival analysis is the cornerstone of many healthcare applications, in which the survival probability from recorded survival times (e.g., time free from a certain disease, time to death) is computed to guide clinical decisions. However, sharing the outcomes of survival analysis may pose privacy concerns, as an adversary may infer whether a target patient is in the data. We quantified this privacy risk and developed a privacy protection method for the Kaplan-Meier survival model.
Learning Objective: -Understand the privacy risk in sharing temporal aggregated data (i.e., results from survival analysis).
-Learn the challenges in modeling potential privacy risk in reconstructing the temporal information associated with the time-to-events.
-Formulate an approach to protect privacy in survival analysis using the notion of differential privacy.
Luca Bonomi (Presenter)
University of California San Diego
Xiaoqian Jiang, UT-Health
Lucila Ohno-Machado, University of California San Diego