Multi-Analyst Differential Privacy for Online Query Answering

Most differentially private mechanisms are designed for the use of a single analyst. In reality, however, there often multiple stakeholders with different and possibly conflicting priorities that must share same privacy loss budget. This motivates problem equitable budget-sharing multi-analyst differential privacy. Our previous work defined desiderata any mechanism in this space should satisfy introduced methods offline case where queries known advance. We extend our on query answering to online answering, come one at time be answered without knowledge following queries. demonstrate unknown ordering results fundamental limit number can while satisfying desiderata. response, we develop two mechanisms, which satisfies all cases but is subject limitations, another randomizes input order ensuring existing