Verifiable Multi-party Computation with Perfectly Private Audit Trail

We propose an efficient protocol for the evaluation of functions getting their inputs from multiple parties in a way that guarantees the result correctness. In our setting, a worker is trusted with the confidentiality of the inputs and, given this assumption, our protocol guarantees perfect privacy to the clients. Our protocol offers an interesting middle ground between traditional verifiable computation protocols, that usually do not come with privacy guarantees and focus on one or a small number of clients, and secure multi-party computation protocol that distribute the privacy trust between a number of parties, at the cost of much more expensive protocols (especially for NP functions and functions that do not admit an efficient static circuit representation) and a demanding infrastructure of independently managed servers interacting in multiple rounds. By contrast, our protocol is single-pass: the clients submit their inputs asynchronously, and everyone can collect the result at any later time. We present three unrelated applications of our technique: solving a system of linear equations, an auction scheme and the search of the shortest path in a shared graph. These examples illustrate the ease of use and the advantage in terms of complexity of our approach. We made a prototype implementation that illustrates the practicality of our solution.