Master-Thesis

Mutually distrusting parties can evaluate a function on their private data using secure multi-party computation protocols. In the standard multi-party computation setting we want to achieve a correct result, while not revealing anything about the inputs. In this work we consider the case that some parties have goals beyond correctness and privacy and therefore might not follow the protocol honestly.

Non-cooperative computation considers players that wish to gain as much information from the execution of a protocol as possible, while preventing others from learning anything at all. In this work we formally introduce the goals of such players. We show when it is an optimal strategy for all players to input correct data and believe the return value. We generalize the existing results on non-cooperatively computable Boolean functions by presenting a theorem for non-cooperatively computable functions for general functions.