1: \begin{abstract}

2: This work considers survival estimation with censored data under setups that preserve individual privacy. We provide an $\alpha$-locally differentially private mechanism on failure indicators and propose a non-parametric kernel estimator for the cumulative hazard function. Under mild conditions, we also prove lowers bounds on the minimax rates of convergence and show that our estimator is minimax optimal under well-chosen bandwidths. The method is illustrated with numerical results on synthetic data.

3: \end{abstract}

4: