The authors propose three differentially private algorithms for constructing confidence intervals for population proportions under stratified random sampling, providing privacy guarantees and asymptotic coverage properties.
