Voting machines (was The Internet of criminal things)

Posted Sep 24, 2015 13:48 UTC (Thu) by dskoll (subscriber, #1630)
In reply to: The Internet of criminal things by marcH
Parent article: The Internet of criminal things

Voting machines are a really terrible solution to something that isn't even a problem. Manual vote-tallying works just fine and scales extremely well; you only need O(log N) tallyers if you use a tree structure.

Voting machines (was The Internet of criminal things)

Posted Sep 24, 2015 20:00 UTC (Thu) by tzafrir (subscriber, #11501) [Link] (10 responses)

The bottleneck is counting the votes. Not sending the results. With a voting machine you have the results immediately. With a manual vote, it takes several hours. And then you need to securely get the ballots physically to a safe place for keeping (to allow re-count).

Voting machines (was The Internet of criminal things)

Posted Sep 24, 2015 20:06 UTC (Thu) by marcH (subscriber, #57642) [Link]

> With a voting machine you have the results immediately. With a manual vote, it takes several hours.

Yeah it's very much like "breaking" news: must choose between getting wrong information immediately versus waiting for some time.

Voting machines (was The Internet of criminal things)

Posted Sep 24, 2015 20:13 UTC (Thu) by marcH (subscriber, #57642) [Link]

> And then you need to securely get the ballots physically to a safe place for keeping (to allow re-count).

With voting machines you: - either have to do the same thing in a possibly more technical and demanding way; - or you don't even have anything at all you can re-count.

From an counting mistake perspective ballot boxes are an "analog" system, meaning errors are negligible and don't matter except for the rare cases where the vote is very tight.

Voting machines (was The Internet of criminal things)

Posted Sep 24, 2015 20:41 UTC (Thu) by raven667 (subscriber, #5198) [Link] (2 responses)

> With a manual vote, it takes several hours.

What is the actual requirement for speed here? For example in the US Presidential election, the ballots are cast on November 8 but the results aren't absolutely required until inauguration on January 20 of the next year, a few hours (or days or weeks) in a 73 day window is not significant.

Voting machines (was The Internet of criminal things)

Posted Sep 25, 2015 3:51 UTC (Fri) by edgewood (subscriber, #1123) [Link] (1 responses)

Well, the Electoral College meets on the first Monday after the second Wednesday in December, which means somewhere between Dec 13 (Dec 1 is a Wednesday) and Dec 19 (Dec 1 is a Thursday). So the results will be needed by then at the latest.

But the bigger problem is that the US has a lot more separate elections per ballot than I think other places with parliamentary systems do. See http://v.gd/7cIVRX for the *front* of a recent election in North Carolina. There are 14 elections on the *front* of the ballot. I couldn't find a good image of the back, but I recall that many or more there.

Voting machines (was The Internet of criminal things)

Posted Sep 26, 2015 15:46 UTC (Sat) by marcH (subscriber, #57642) [Link]

> But the bigger problem is that the US has a lot more separate elections per ballot than I think other places with parliamentary systems do

Maybe an interesting trade-off could be "hardware assisted tallying", where a scanner operates slowly enough that anyone in the room can keep an eye on it.

Voting machines (was The Internet of criminal things)

Posted Sep 25, 2015 2:18 UTC (Fri) by bfields (subscriber, #19510) [Link]

I don't think it's even necessary to choose. E.g. here in Michigan the ballot is a machine-readable paper form that you fill out and then feed into a scanner built in to the top of the ballot box. The scanner tallies the results and then stores the ballot inside. At the end of the day the poll workers can report instant (preliminary) results, but the ballots can also be recounted manually later to confirm.

(For some reason, I don't think that manual recount necessarily happens in the absence of a challenge. It's easy enough to do.)

You fill out the form by hand with a pencil. I haven't used one, but I seem to recall being told that the polling places also have electronic voting machines, but that all they do is produce paper ballots. That can be helpful for voters that have some disabilities (poor eyesight, or whatever). I suppose in theory you could compromise those but I think that would be easier to detect (and less rewarding) than a similar attack on a purely electronic system.

Voting machines (was The Internet of criminal things)

Posted Sep 25, 2015 12:25 UTC (Fri) by NAR (subscriber, #1313) [Link] (2 responses)

I think at the last Hungarian elections the counting was finished in 3 hours in the first round (where everybody had two votes) and maybe 1,5 hours in the second round (where voters had only a single vote). The political analyzers on TV barely had the time to speculate before the results were ready.

So I don't think counting is a bottleneck currently. However, if we'd have more complicated voting system like the Schulze method, then having the votes in a computer could be more useful.

Voting machines (was The Internet of criminal things)

Posted Sep 29, 2015 16:05 UTC (Tue) by martin.langhoff (subscriber, #61417) [Link] (1 responses)

Agreed - many countries get same-day results with paper votes, most notably the EU countries and NZ.

In the USA there's been some odd media fabrication of "it can't be done unless we replace the paper vote with horridly unreliable computers". It is clearly not true, but it would take looking at how other countries do things... so... ;-)

Voting machines (was The Internet of criminal things)

Posted Sep 29, 2015 16:42 UTC (Tue) by marcH (subscriber, #57642) [Link]

You meant... "inferior" countries? :-)

Conversely: http://arstechnica.com/business/2014/08/chip-based-credit...

Voting machines (was The Internet of criminal things)

Posted Sep 27, 2015 18:57 UTC (Sun) by debacle (subscriber, #7114) [Link]

Counting might even take days. But as long as the election period is much longer, it doesn't matter at all.

Voting machines (was The Internet of criminal things)

Posted Sep 25, 2015 7:44 UTC (Fri) by jezuch (subscriber, #52988) [Link]

> Voting machines are a really terrible solution to something that isn't even a problem.

To be fair, the current crop of voting machines is a terribly naive implementation of the concept, wasting an enormous opportunity for improvement. They basically do nothing more than the human tallyers can do, except faster - and with less auditability. But there are algorithms that [promise to][1] allow the vote to be both secret and auditable, by anyone who can count, not just the tallyers. But to do that you need to implement actual cryptography, not the half-arsed "add 1 to this variable" machines we are offered today.

[1] I'm not a cryptographer so me not being able to spot any flaws in such an algorithm doesn't mean anything, obviously :) Anything that wants to be used in actual elections need to be first exposed to intense scrutiny by the security community. Which is he opposite of the current state of affairs.