As we all know, secure multi-party computation allows us to run a certain computation/function on private inputs contributed by different parties (that do not necessarily trust each other). The security guarantee of an MPC is that parties learn nothing beyond the computation result.
However, MPC does not deal with the question of whether the function (output) itself reveals much information about the inputs.
Question: Is there any class of functions that ensures an adversary (given the output) cannot learn anything about the honest parties inputs?