Achieving Differential Privacy in Secure Multiparty Data Aggregation Protocols on Star Networks
We consider the problem of privacy-preserving data aggregation in a star network topology, i.e., several untrusting participants con- nected to a single aggregator. We require that the participants do not discover each others data, and the service provider remains oblivious to each participants individual contribution. Further- more, the final result is to be published in a differentially private manner, i.e., the result should not reveal the contribution of any single participant to a (possibly external) adversary who knows the contributions of all other participants. In other words, we require a secure multiparty computation protocol that also incorporates a differentially private mechanism. Previous solutions have resorted to caveats such as postulating a trusted dealer to distribute keys to the participants, or introducing additional entities to withhold the decryption key from the aggrega- tor, or relaxing the star topology by allowing pairwise communica- tion amongst the participants. In this paper, we show how to obtain a noisy (differentially private) aggregation result using Shamir se- cret sharing and additively homomorphic encryption without these mitigating assumptions. More importantly, while we assume semi- honest participants, we allow the aggregator to be stronger than semi-honest, specifically in the sense that he can try to reduce the noise in the differentially private result. To respect the differential privacy requirement, collusions of mu- tually untrusting entities need to be analyzed differently from tradi- tional secure multiparty computation: It is not sufficient that such collusions do not reveal the data of honest participants; we must also ensure that the colluding entities cannot undermine differen- tial privacy by reducing the amount of noise in the final result. Our protocols avoid this by requiring that no entity neither the ag- gregator nor any participant knows how much noise a participant contributes to the final result. We also ensure that if a cheating ag- gregator tries to influence the noise term in the differentially private output, he can be detected with overwhelming probability.