Problem statement
Stub
MACI, minimal anti-collusion infrastructure is a voting system, suggested here: https://ethresear.ch/t/minimal-anti-collusion-infrastructure/5413 by Vitalik Buterin, and implemented here: https://github.com/privacy-scaling-explorations/maci.
We are aiming to improve it by swapping a singular server for an MPC in such a way that user privacy could be broken only by collusion of parties in MPC.
Other desired properties - non-coercibility and auditability should still hold.
We mainly consider some similar scenarios:
- Third party is trying to force user into voting as they please.
- Third party is trying to waste user's vote.
- Third party is trying to buy user's vote.
- User themselves is trying to sell their vote.
From a game-theoretic standpoint, they are all essentially equivalent, for example coercion can be treated as a particular case of vote-buying, where attacker agrees to not harm user in exchange for the vote.
We assume that third party doesn't have an ability to monitor every user's action (for example, that requires that user unconditionally posesses the key and is not locked in a trusted environment), but they can force them to send transactions of desired form.
Actually, attacker can even ask user to blindly sign transactions with unknown content via engaging in 2-PC protocol - so effectively it should be thought that attacker can also access user's key (and crypto-economic defenses from key-selling are, thus, ineffective).
It should also be that, similar to normal MACI, we assume that attackers are generally trying to coerce voters in a specific vote, not trying to purchase voting rights from them fully. Formally, it means that there is a first round, where user can not interact with the attacker.