Efficient Verification of Input Consistency in Server-Assisted Secure Function Evaluation

We consider generic secure computation in the setting where a semi-honest server assists malicious clients in performing multiple secure two-party evaluations (SFE). We present practical schemes secure in the above model. The main technical difficulty that we address is efficiently ensuring input consistency of the malicious players across multiple executions. That is, we show how any player can prove he is using the same input he had used in another execution. We discuss applications of our solution, such as online profile matching.