“In this work, we study the problem of privately maximizing a submodular function in the streaming setting. … When the size of the data stream drawn from the domain of the objective function is large or arrives very fast, one must privately optimize the objective within the constraints of the streaming setting. We establish fundamental differentially private baselines for this problem and then derive better trade-offs between privacy and utility for the special case of decomposable submodular functions.”

Find the paper and full list of authors in Proceedings of Machine Learning Research.