Thursday, November 02, 2006

An Open Letter to Jay Nordlinger

(This is in response to Jay Nordlinger's column, in which he complains that if Democrats win, problems with voting machines will not be followed up.))

Though I don't doubt you're correct that the media won't pay attention to voting machine irregularities if Democrats win, that doesn't mean they're not worth paying attention to even if Republicans win. The right to a fair vote is fundamental to our democracy, and if it's abrogated, it will most probably encourage even more people to not vote. I fear that too many people, after reading your column, will dismiss any reports of voting problems as partisan bickering instead of evidence of a real problem with electronic voting machines. Here are a few problems that have already been uncovered with electronic voting machines so far this election season:

  • In Cuyahoga County, Ohio, the county Board of Elections commissioned two separate investigations into problems they had during the May 2nd primary. They
    found that the voting machines weren't even internally consistent-- the voter-verified paper audit trail reported one set of votes, the summary printed at the end reported a different set, the memory cards used by the machines reported a third, and the election archive reported a fourth! That's four different totals, from one machine! How can we possibly trust our democratic process to voting machines that can't even agree with themselves?

  • Also from Cuyahoga County: most people aren't aware of this, but it's crucially important that the voter-verified paper trails mentioned above must be marked so that they can be checked against the machine they were drawn from. Otherwise, they're not a meaningful check. But poll worker were either poorly trained or provided with the wrong equipment, because many paper trails were stuffed in random, unlabeled canisters, rendering their crucial use as a check on unauditable databases (see above) meaningless.

  • Again from Cuyahoga (this is the last one, I promise!): after the primary election, 29 machines just vanished. *poof* A few officials took them home with them, but most simply dissapeared. You are likely not aware of this, but a team at Princeton was able to develop a vote-stealing computer virus based on having *one* machine at their disposal-- no source code, no manuals, no documentation of any kind, just the machine. Somebody has enough machines to run an entire precinct election; maybe they don't intend fraud, but there simply aren't any controls in place to prevent that.

  • In Texas and Florida, people have already reported that although they pushed the button on the screen to vote Democrat, the screen showed a Republican being elected. Even though poll workers were called over to help, nobody could figure out how to stop this from happening, so the person in question had to go ahead and vote for the wrong people-- or did she? (See Cuyahoga County, above.)

  • Although Diebold gets a lot of grief over their machines, other vendors are not immune from problems. A voter can vote an arbitrary number of times on a Sequoia voting machine using only a Post-It note. This is not a theoretical vulnerability, this is something easy to do with a pen and a Post-It. I don't know about you, but this scares the pants off me-- imagine what the old Mayor Daley could have done with technology like this!

There are even more examples at Ars Technica if you're at all interested, and I pray and hope that somebody out there is. The root of all these problems is that electronic voting machines are, at bottom, networked computers, and most election boards don't understand this. This is not to their discredit-- they weren't sold a bunch of networked voting computers, they were sold independent machines. They were also under the gun of the Help Americans Vote Act, which encouraged them to buy what was not even beta-tested software, and after they discovered this (and most of them have by now), they had spent so much time and money training volunteers that there was no way they could go back to a non-electronic solution.

As a computer scientist, if somebody asks me if I voted, all I can tell them, at this point is, "I think so." I don't consider it a nice thing if I can say "yes" after the next election; I consider it essential.