I need to know in an outsourced two party computation where honst $A$ and $B$ outsource their private and secure data to a malicious server, why we need to design a simulator that interacts with an adversary who exists in real model?
See the paper cited below as an example:
