A1552
Title: Debiasing differentially private time-to-event data
Authors: Yi Xiong - University at Buffalo (United States) [presenting]
Abstract: Sharing time-to-event data is essential for enabling collaborative research, designing effective interventions, and advancing patient care. However, sharing exact survival curves poses significant privacy risks. To protect individual privacy and mitigate the risk of membership inference attacks, various privacy-preserving solutions have been proposed. The differential privacy (DP) framework, in particular, offers strong and rigorous protection for data sharing. However, the noise injection required by DP can distort the probability density of the data, resulting in low utility and invalid statistical inference. Methods are proposed to mitigate these biases in differentially private, right-censored time-to-event data using a deconvolution framework via kernel density estimation. A bias-corrected nonparametric estimator for the marginal distribution of event times is provided and a corrected score approach for regression analysis is developed to ensure valid inference under privacy-preserving noise.
