Differential privacy, equilibrium, and efficient allocation of resources

In this talk, we survey the recently defined notion of joint differential privacy, together with several of its applications. Joint differential privacy is an appropriate solution concept for problems whose solution can be partitioned amongst the n agents who provide the private data to the algorithm: the two examples we focus on are allocation problems and equilibrium computation. Our focus in this survey is twofold: on purely algorithmic problems that can be solved under the constraint of joint differential privacy that cannot be solved under the standard constraint of differential privacy, and on the incentive properties of mechanisms that compute some kind of “equilibrium” under the constraint of joint differential privacy.