Secure multi-party computationSMPC

A way for several parties to work out a shared answer together without any of them seeing the others' private data.

It is like a group finding out whether they share a secret number without anyone saying their own number out loud.

A family of cryptographic methods that let several parties jointly compute a function over their private inputs while keeping each party's input hidden from the others, so that only the agreed result is revealed. Secure multi-party computation is often shortened to SMPC or MPC.

SMPC rests on a few building blocks:

• Secret sharing, in which each input is split into random-looking pieces, called shares, that are spread across the parties, and no single piece reveals anything.
• Computing on shares, in which the parties run the calculation on the pieces and combine the outputs into a final answer.
• Private set intersection, a common use in which parties learn which items they have in common, such as matching identifiers, without revealing the rest of their lists.

A simple example works like this:

1. Two banks each hold a list of identifiers they have flagged.
2. Each bank splits its identifiers into shares, so neither list is visible to anyone.
3. The parties compute on the shares to find whether an identifier appears in both places.
4. They learn the answer, such as match or no match, and nothing else about the other side's data.

In TrustGraph, each query uses fresh, one-time cryptographic values, so the same identifier produces a different query every time, and the participants compute only on cryptographic shares, which prevents anyone from linking queries together.

SMPC sits next to terms that are easy to mix up:

• Encryption protects data while it is stored or sent, while SMPC protects data while it is being used in a calculation.
• De-identification removes or transforms identifiers before data is used, and it can be combined with SMPC.
• Fraud consortium pools data in one place, while SMPC lets parties collaborate without pooling it.