Slashdot Mirror
E-Voting Glitch: 19,000 Voters, 144,000 Votes
nick_davison writes "The Indianapolis Star is reporting the latest case of 'interesting' E-voting results. Tuesday's Boone County election, using MicroVote software returned 144,000 votes from 19,000 registered voters. After much panicking and tracking down the bug, the actual number of votes turned out as 5,352. With yet another mistake, does anyone still trust closed-source electronic voting?"
Personally I like the bit about vote-counting in France. Sounds a lot more advanced (read: secure) than the US way of doing it.
Pen? We use thick pencils, with fairly soft cores, attached to the polling booth by a long piece of string! No change of the ink drying up, and little chance of the pencil breaking.
Are you serious? Are the people who count the votes not volunteers in the US?
Having an extra 100,000+ votes clearly stands out as an error. I would have been more concerned if it was a small enough number not to be detected, but a big enough number to affect close races.
what causes me more worry are the bugs (features?) in these machines that are known only to a select few. i was hoping that after the elections last week more hue and cry would be made in the mainstream media about these machines by the candidates who lost. that doesn't appear to be forthcoming, though. pity.
Your objections are certainly justified; on the other hand Germany where I am living is doing all of its voting the traditionall pen-and-paper-ballot way, and we get first projections minutes after the voting closes, more and more reliable projections shortly after and very accurate (usually 0,x % to the official final results) inofficial final results the same evening (usually our voting booths close at 6 pm). The official results are available IIRC about 2-3 days after the vote.
The people staffing the voting booths and counting the votes are usually volunteers who get a small payment for their troubles. All in all our systems
seems to work quite well.
And even if Germany is far smaller than the US it has still a not too small voting population.
I used to be a hard core political junky.
There is a extremely large amount of vote fraud going on now with the paper ballots, mostly for local elections. (nobody in the big parties talk about it because it would cause too much trouble)
One of the big ideas of computer voting is you remove the ability to add, replace or destroy ballots in the time gap between voting and being tallied.
Open-sourcing the voting software is important, but in my opinion, not as important as maintaining separate systems for ballot printing and ballot tabulation.
I wrote about it in this journal entry.
"Lawyers are for sucks."
- Doug McKenzie
With yet another mistake, does anyone trust electronic voting full stop?
Or as some of the American Electorate might say; "with yet another mistake does anyone trust voting full stop". I think the source of the problem is the perception by various interests in the US that there is some form of money to be made in these systems. This is wrong. Get the _process_ of electronic voting designed right (I mean imagine the first elections back in the year dot. All those who vote for Trevor stand to the left, all those for Dave to the right, all those for Ug, um well, you just stand where you are... No dave, stop killing the people voting for Trevor... What do you mean you don't want to vote for Ug, well ok then you just stand over there... No I don't care who you want to vote for they're not here. Oh fuck it, this is too hard). Then the implementation simply becomes a question of reducing cost. There is no "marginal" profit to be had and as such there is almost no way that private enterprise can fund the development of these systems better than the state. The argument for free software systems is equally persuasive.
Then there is the deployment of the hardware/infrastructure to actually deliver the voting functionality to the electorate (and that is something that can get better and better over time as well). It is very expensive and the only benefits compared to the counting of paper votes are accuracy and cost savings (for get speed, it's not like there is a power vacuum before the result. so what if it takes a few days). If you can give accuracy then get out of the game and the only way to reduce cost is to fund on a cost basis which means the state should fund the system not enterprise.
"The first thing to do when you find yourself in a hole is stop digging."
Comment removed based on user account deletion
RP
It looked like they used this machine to scan it: www.essvote.com
Very clean. The number of votes was called in and double checked against the smart card inside which connects by modem. Results 20 minutes after the polls closed and a paper trail if needed. Great stuff.
Machines crashing while the polls were open
Central collection point jammed with call-in traffic (understandable)
Machine inflates count almost 30 times the actual figure.
Alright, I give up. Let us at least try to put a positive spin on this issue. Were there any elections that didn't have problems when using the new electronic voting systems? And what was the ratio of non-problematic electronic voting to problematic electronic voting? I'd say that if more than half of the electronic voting machines had problems, the manufacturer should be sued. I'd advocate a lawsuit to get out from under any contracts that may exist for the installation and maintenance of this equipment.
An aside: Does anyone know whether or not computer scientists had any input at all on the design of these beasts? If not, then what a terrible waste of good talent. I don't know, maybe I'm wrong there, because I still think an electronic voting machine wouldn't be very complicated to design.
Fred
"A fool and his freedom are soon parted"
-RMS
Vote-counting is up there with life support systems in terms of how critical mucking it up can be. If you need to have three independently written programs doing the counting and comparing results, then do it. For something as simple as this, I disagree that common mistakes are acceptable.
May we never see th
Fortunately, MicroVote is an also-ran (Loser, in political terms) with a handful of customers in a handful of states within a couple hundred miles of where they're based. Boone County is probably now thanking the wisdom of choosing an unproven local company for such a "simple" task.
Speaking as an industry insider, though, counting votes really isn't simple. It's damn tough. In about 99.999% of the screw-ups you hear about, the ultimate cause is human error on the part of election workers. Any computer can take an int (which starts out as a correct count) and ++ it correctly pretty much 100% of the time.
Here's one example of a problem inherent in counting votes, though. You can't count 5 million paper ballots on one machine capable of 300 ballots/minute. So you use multiple machines, and you have to combine results. Any time you combine results, you have users who occasionally forget to add the totals from one machine or mistakenly add the totals from another machine twice. Sure, you can build in a safeguard that indicates an error if results from one machine are brought in a second time, but it's about as effective as "Are you sure you want to delete this file?" People want updates every 30 minutes, they want projections...
Forget about Open Source vs. Proprietary. It's hard enough to get things right when you're trying. There are more opportunities for error than any of you probably realize. That's why there is extensive testing with sample ballots in the days before an actual election to verify that the machines are counting properly.
You start with defining all the election information -- did you spell that candidate's name correctly? Was she classified with the correct party? Is that contest appropriate for these precincts, or is it only for the 25 Green Party members in this neighborhood? Don't forget to print each 1/4 of the ballots with a different ordering of the 4 candidates so as to avoid positional bias! I've seen 100 different ballot styles used in a precinct with only a few thousand registered voters...and then the counts must be combined up to a national contest level.
You have potential PostScript errors, printer errors with ballots printed out of spec for the scanners, too few ballots printed, poll workers not showing up, ballots soaked by rain, ballots misplaced by inexperienced volunteers...
And then you have to deal with a candidate either added to the ballot or removed from the ballot at the last minute (literally!). You've already gone through the lengthy, week-long process of validating the counting of all the different ballots in different machines, and now you have 24 hours to change something.
Actually, the best is when some judge decides two days before an election that the way some votes are counted in certain circumstances is illegal and must be changed. The machine firmware was independently examined, tested, and certified months in advance, and now it must be changed with hours to go before an actual election. There's no arguing with the Judicial System!
So tell me counting votes accurately and reliably is simple. I think I'll ignore you and keep working on removing all the instances of "Bush" hardcoded in this tabulator firmware... (that's a fscking joke!)
The main thing is that there should always be a paper receipt as backup. When you go to the ATM you get a receipt, when you use your credit card, you get a receipt. When you vote electronically, a matching receipt should be printed, signed by the voter, and retained in a locked ballot box. The receipts in the ballot box can then be counted if there is a question about the electronic results.
I think we need to consider keeping the ability to match voters to ballots in order to reduce the chance of ballot-box stuffing (either electronic or physical). Of course safeguards would need to put in place to restrict and prevent the abuse of knowledge about how someone voted. For example, after a certain amount of time, all ballots should be destroyed, etc.
I'd trust closed-source voting before I'd trust open-source where a potential exploiter can easliy scan the code finding and exploiting bugs. Though I don't see how it can be so hard to design a secure voting system. You wouldn't want it to run on anything more than the most basic kernel though, so the best idea is probably to use a simple proprietary OS. And you could architect the thing so that it only runs some sort of registered instructions. Also, the whole thing would have to be closed-circuit obviously.
To my mind, the problem that these computers were meant to solve was the production of legible, non-ambiguous, easily tallied ballots that accurately reflect the voter's intentions in the booth.
I see a computer terminal that is very straightforward and relatively low tech. All this terminal does is display the choices, record the user's input, and spit out a chit with the voter's choices displayed in human and machine readable form. These votes could easily be placed through a bubble reader or cross-checked by humans.
This is tech people can understand and verify on the spot before they cast their ballot into the box. Is there really any reason to have that terminal record the votes, tally the votes, and wire the totals to centralized servers? How many points of failure/corruption do we really want here?
"The world is a construct of forceful imagination. Those who don't know walk around in the reailties of those who do"