Secure Statistical Analysis of Distributed Databases, Emphasizing What We Don't Know

Abstract

Over the past several years, the National Institute of Statistical Sciences (NISS) has developed methodology to perform statistical analyses that, in effect, integrate data in multiple, distributed databases, but without literally bringing the data together in one place. In this paper, we summarize that research, but focus on issues that are not understood. These include inability to perform exploratory analyses and visualizations, protections against dishonest participants, inequities between database owners and lack of measures of risk and utility.