Multiparty Reach and Frequency Histogram: Private, Secure, and Practical

Abstract

Consider the setting where multiple parties each hold a multiset of users and the task is to estimate the reach (i.e., the number of distinct users appearing across all parties) and the frequency histogram (i.e., fraction of users appearing a given number of times across all parties). In this work we introduce a new sketch for this task, based on an exponentially distributed counting Bloom filter. We combine this sketch with a communication-efficient multi-party protocol to solve the task in the multi-worker setting. Our protocol exhibits both differential privacy and security guarantees in the honest-but-curious model and in the presence of large subsets of colluding workers; furthermore, its reach and frequency histogram estimates have a provably small error. Finally, we show the practicality of the protocol by evaluating it on internet-scale audiences.

An Effective and Differentially Private Protocol for Secure Distributed Cardinality Estimation

“…However, the protocol is not differential private (refer to Section 3.5 for details) when the DHs are honest but curious. Ghazi et al [4] proposed to estimate both the union set's cardinality and the frequency histogram (i.e., the fraction of elements appearing a given number of times across all the DHs). Their protocol is based on a variant of Bloom Filter and consists of multiple phases with noise injected to guarantee differential privacy.…”

“…On the Internet, an advertiser often conducts a campaign across several publishers, which show advertisements on behalf of the advertiser. The campaign's reach is defined as the number of distinct users exposed to the campaign by at least one publisher, which is a critical metric for evaluating the campaign's efficacy [3] [4]. Different publishers may reach overlapping sets of individuals.…”