Identifying Adversaries

The goal of multi-party computation (MPC) is to compute a function taking multiple parties’ private inputs without revealing those inputs to other parties. In general, if a majority of parties are controlled by a malicious adversary, the best we can achieve is “security with abort” where adversaries can always force the protocol to abort.

In an “Identifiable” MPC (IDMPC) protocol, honest parties can agree on the identity of some adversary any time the protocol aborts, allowing them to remove that party and start over.  A recent work successfully reduced $n$-party IDMPC to $n-1$-party IDMPC in the case of constantly many honest parties. This project extends these results to any number of honest parties.