Friday, June 13, 2008

So Now Can We Talk About Hand Counts?

I wanted to share yesterday's opening comment by John Gideon in Voters Unite's Daily Voting News because it refers to some important work by John Washburn that I may have played a small role in motivating. I'm also going to report on the state of play in NY, the last state to resist implementation of so-called "HAVA-compliant" voting systems in favor of retaining its sometimes imperfect but usually reliable mechanical lever machines. (The levers are also HAVA-compliant as long as HAVA Accessibility requirements are met with other equipment, but this has yet to be properly adjudicated.)

Gideon writes:

Voting systems are tested for federal certification by a "Voting Systems Test Lab" (VSTL) against the voting systems standards. So, one would expect, that if a voting system fails to meet the standards it fails testing and cannot be certified. One would expect. However, the voting systems standards give the vendors and the VSTLs an out. B.5 of Appendix B of Volume II of the 2002 VSS and the 2005 VVSG says, “….any uncorrected deficiency that does not involve the loss or corruption of voting data shall not necessarily be cause for rejection”. A Premier voting system was recently recommended for certification by Systest, one of the VSTLs. In the test report Systest reported 79 discrepancies. 77 of those discrepancies were serious but fell within the B.5 “get out of jail free card”. Two of the 79 were serious discrepancies that report loss of data and are not covered by the “get out of jail free card”. Yet, the system was recommended for certification and the EAC is considering following that recommendation.
Well, I hate to say "I told you so", but thanks to John Washburn for pointing out:
"[B]oth sets of [2002 and 2005 federal voting system] standards have an explicit loophole that allows almost all the requirements — weak as they are — to be ignored. This second objection was first brought to my attention two years ago by Howard Stanislevic."
It's always nice to see that someone else has been paying attention!

As for how the testing in NY is going, so far, it makes the above discrepancies uncovered by Washburn look like a walk in Central Park!

Thanks to the Honorable Gary L. Sharpe, the judge in the case of US v. The New York State Board of Elections who has previously stated that he gets his information about electronic vote counting from reading the newspapers, the State Board of Elections is now required to send the Court weekly status reports on the progress of, among other things, lever machine replacement testing, also known as "Plan A." According to the report dated June 6, 2008:
SysTest reports that Sequoia/Dominion has 279 open source code discrepancies and ES&S has [a whopping] 915 open source code discrepancies.
This, even according to the inadequate 2005 Voluntary Voting System Guidelines that John Washburn, myself, and others such as UC Berkeley computer science professor Dr. David Wagner have been criticizing for years.

But with the loopholes reported above and the pressure from the US Dept. of Justice and the Court, what assurances will NY voters have that any of the above standards violations will ever be corrected?

Not that source code is the only way to screw up an election. Far from it! Those of us actually paying attention know that ballot programming, also known as election configuration, is much more accessible and dangerous than mere source code. Anyone with access to an Election Management System such as GEMS, Unity or WinEDS (the big 3) already has all the tools necessary to manipulate election results, even BEFORE the election, by tinkering with ballot programming. But voting system source code certainly has the potential to do just as much damage if it doesn't work the way it's supposed to after it's complied and run a voting system.

As New York prepares to dismantle its lever voting system (which, for some uninformed folks, just can't happen soon enough), the fact is to date, there is still no suitable replacement available. With a total of over 1,000 standards violations, even based on the weak Federal standards, it's hard to imagine how they can all be corrected in time for a 2009 election (2008 is already officially out of the question, except for deployment of electronic ballot markers to comply with HAVA's well-motivated Accessibility requirements).

It's the vote counting (stupid)! That's what needs to be checked by counting enough ballots by hand to see who really won our elections. To that end, the State Board of Elections have proposed some new election auditing regulations. Unfortunately, they are still inadequate, but they can be salvaged if the Board would exercise some due diligence and consult with those who, for years, have been studying the problem of confirming electoral outcomes without having to depend on source code, ballot programming or election configuration and management software. Maybe they will.

No comments: