Usable and Efficient Secure Multiparty Computation

With the aim of implementing a privacy preserving distributed monitor framework, we identified a scheduling strategy for a distributed algorithm. The algorithm assumes a set of primitives (i.e. language intersection) that can be executed by two parties preserving their privacy. On top of these primitive interactions, we verified that our scheduling policy does not leak information if the participant topology graph (i.e. the graph where two participants are directly connected if the intersection of their alphabets is not empty) is a tree.

To enable the algorithm we started the implementation of two party language composition. In particular, we focused on allowing to privately compute the intersection of the party languages. If the languages are regular (i.e. the ones that can be expressed using regular expressions), the natural approach is to compute the minimization of the synchronous product of the deterministic finite automation that represent the languages. We are investigating the effectiveness of the existing minimization algorithms (in particular Hopcroft and Brzozowski) when applied to our case studies and implemented on the Sharemind platform. We are also exploring if we can constraint the input languages to reduce the cost of the computation.