Privacy-Preserving Multi-Party Clustering: An Empirical Study

Enterprises are transitioning towards data-driven business processes. There are numerous situations where multiple parties would like to share data towards a common goal, if it were possible to simultaneously protect the privacy and security of the individuals and organizations described in the data. Motivated by the increasing demands for data privacy, this paper provides the first comprehensive evaluation of privacy-preserving multi-party computation. As a case study, we consider the clustering task, which consists of grouping a set of points based on their similarity. Our goal is to understand the trade-offs involved when different parties want to collaboratively perform a computation while preserving their data privacy. In particular, we study the implications of centralized and distributed privacy-preserving solutions (such as encryption and data perturbation) on clustering quality, privacy and computational performance. Our results offer a new perspective on multi-party computation for both service providers and users, highlighting the drawbacks of these approaches and opening opportunities for future research.