Distance Computation between Two Private Preference Functions

Abstract : We consider the following problem: two parties have each a private function, for example one that outputs the party’s preferences on a set of alternatives; they wish to compute the distance between their functions without any of the parties revealing its function to the other. The above problem is extremely important in the context of social, political or business networks, whenever one wishes to find friends or partners with similar interests without having to disclose one’s interests to everyone. We provide protocols that solve the above problem for several types of functions. Experimental work demonstrates that privacy preservation does not significantly distort the computed distances.