d3p - A Python Package for Differentially-Private Probabilistic Programming

Abstract We present d3p , a software package designed to help fielding runtime efficient widely-applicable Bayesian inference under differential privacy guarantees. achieves general applicability wide range of probabilistic modelling problems by implementing the differentially private variational algorithm, allowing users fit any parametric model with differentiable density function. adopts programming paradigm as powerful way for user flexibly define such models. demonstrate use our on hierarchical logistic regression example, showing expressiveness approach well ease running parameter inference. also perform an empirical evaluation complex and find ~10 fold speed-up compared implementation using TensorFlow Privacy.