Wednesday, September 27, 2006

A new voting scheme

Ron Rivest has just put out a voting scheme that is remarkable for its simplicity. It addresses the following questions:
  • How do you tell if your vote was counted
  • How might you get proof that you voted
  • How can this be concealed from (a) someone trying to coerce you (b) someone trying to find out your vote or tamper with it (c) from you !
I'm not an expert on voting schemes, so I'll leave the detailed analysis to the experts. I will note that the scheme is NOT cryptographic, and is surprisingly elegant in its basic method. One major drawback: it requires each voter to vote on 3 ballots in a specific way: in the era of hanging chads and butterfly ballots, this might be viewed as a major drawback.

Here's the idea: suppose you have to vote on one of two candidates Alice and Bob. You vote three times, once on each of three ballots. If you prefer Alice, you vote for her on two of the three ballots, and vote for Bob on the remaining ballot. As a receipt, you take back a copy of one of the three ballots you used. The three ballots are separated and cast individually as separate ballots into the counting box.

The neat idea here is that each candidate gets n + C votes, where C is the number of people who voted for them. So it's still easy to determine winners, and because the single receipt ballot could have come from any combination of votes, the true intention of the voter cannot be determined from their receipt. There is some heuristic reasoning about the possibility of tampering and where the main weaknesses lie.

An important side effect of the fact that you can't determine a vote from a receipt is that vote selling is not possible: how will you prove to the vote buyer that you voted as they asked ?

(From comp.risks via Adam Buchsbaum)


No comments:

Post a Comment

Disqus for The Geomblog