Reliable Intersection Computation within Malicious Environments

Abstract

A secure calculation of common data (D 1 ∩ ... ∩ D n ) of different participants without disclosing D i is useful for many applications and has been studied as the Secure Multiparty Computation problem. However, proposed solutions assume all participants act “semi-honest”, which means participants may neither alter the protocol execution nor fake database content. In this contribution, we focus on malicious participant behavior and prove that an atomic exchange of common data is not possible under the assumption of malicious participants. We propose a mechanism to calculate the intersection of multiple participants, which does not only reduce the disclosure in case participants cheat by altering the protocol to a negligible amount, it is also resistant against malicious participants that cooperate in order to cheat others. Furthermore, it impedes database content faking, which could be done when using other protocols by participants in order to check if data is contained in the other’s databases. Last, we show experimentally the practical usability of our protocol and how the level of trust has an impact on the exchange speed of the intersection.