This entire thread is interesting because our typical programming instincts - making sure user actions are linked to user desires via digital signatures, etc - get completely thrown out the window when you talk about voting and secret ballots. You need to be able to ensure the voter is able to make their choice independently, without pressure or publicity - but you can not perform any kind of integrity check that would link the vote back to the voter.
Nicely said. I have been pondering the comments and wondering if there is an over riding problem. From my (less than perfect) understanding of the US electoral system, it is possible to have a president elected who has been voted for by considerably less than half of valid votes. This part of the systems seems more broken to a non-US citizen than a (hopefully) single voting machine being caught on camera breaking.
Nice to see Civitas would use a tamper evident log file (rolling temporal hash). Alas, generally, encoding the order of the ballots cast destroys voter privacy.
I stand by my earlier comment (cross thread): These crypto based voting systems rely on hash collisions to hide individual ballots within a herd of ballots. Because Civitas encodes votes as ranked preference (to support winner takes all, Condorcet, approval voting), there's even more information contained within each ballot, decreasing the likelihood of a hash collision, increasing the likelihood of inferring each voter's unique ballot.
Something did occur to me, however. Right now, all races are encoded onto a single ballot. Making it more likely that every ballot within a precinct is utterly unique.
But if each race was split onto its own ballot, then a crypto based voting system might be workable.
Hmmm.
As loathe as I am to validate a crypto-based scheme in any way, these schemes aren't going away, no small part because the geeks keep pushing technological fixes for perceived societal problems. So I'm somewhat resigned that I should make the most of it, help make sure the worst parts are mitigated.
