Beyond Cryptography: Deniable Privacy for Secure Data Aggregation

Abstract

We assess the privacy properties of the count function, an essential data aggregation primitive, in the context of a real-world secure data aggregation platform called SCRAM (Secure Cyber Risk Aggregation and Measurement). Subject to the constraints of few data contributors and a limited tolerance for noise in the output of the count function, we seek an alternative to differential privacy, and we develop a new privacy-preserving mechanism called deniable privacy. We show that deniable privacy provides the proper balance between accuracy and privacy in the case of SCRAM, and we demonstrate that the utility of deniable privacy extends broadly to other data aggregation applications.