Private Sequential Learning

Can we learn privately and efficiently through sequential interactions? A private learning model is formulated to study an intrinsic tradeoff between privacy query complexity in learning. The formulation involves a learner who aims scalar value by sequentially querying external database receiving binary responses. In the meantime, adversary observes learner’s queries, although not responses, tries infer from them of interest. objective obtain accurate estimate using only small number queries while simultaneously protecting his or her making provably difficult for adversary. main results provide tight upper lower bounds on as function desired levels estimation accuracy. authors also construct explicit strategies whose optimal up additive constant.