Privacy-preserving Cooperative Statistical Analysis

Wenliang Du, Mikhail Atallah
Purdue University

The growth of the Internet opens up tremendous opportunities for cooperative computation, where the answer depends on the private inputs of separate entities. Sometimes these computations may occur between mutually untrusting entities. The problem is trivial if the context allows the conduct of these computations by a trusted entity that would know the inputs from all the participants; however if the context disallows this then the techniques of secure multi-party computation become very relevant and can provide useful solutions.

Statistic analysis is a widely used computation in real life, but the known methods usually require one to know the whole data set; little work has been conducted to investigate how statistical analysis could be performed in a cooperative environment, where the participants want to conduct statistical analysis on the joint data set, but each participant is concerned about the confidentiality of its own data. In this paper we have developed protocols for conducting the statistic analysis in such kind of cooperative environment based on a data perturbation technique and cryptography primitives.

Keywords: Privacy, secure multi-party computation, statistical analysis

Read Paper Read Paper (in PDF)