Allegheny County's audit of 18 touch-screen voting machines this week was a good first step, advocacy group VoteAllegheny said, but it fell short of being complete.
"When you have zero confidence and you get any improvement, that is a giant boost," said David Eckhardt, a computer science professor at Carnegie Mellon University and a member of the group.
At a county board of elections meeting yesterday, VoteAllegheny members commended the county for doing the audit, carried out over a five-day period by SysTest Labs, a Denver-based software verification firm.
"This [testing] was magnificent and thank you for your leadership," said Richard King, a VoteAllegheny member.
The group, which for months has asked the county to verify the authenticity of the software loaded in the voting machines, had proposed that the county do a random-sample audit of about 40 of its 4,700 machines.
The county instead audited 18 machines because of time constraints and the fact that the goals of the audit were achieved by the sample of machines that were tested, said Geoffrey Pollich, SysTest director of state compliance consulting.
"The first eight machines were used as a sample to develop the audit protocol. We then chose 10 machines randomly over the warehouse and we started running the test," said Mr. Pollich, whose company was recommended to conduct the audit by the state.
SysTest is the company that verified software in the touch-screen machines manufactured by Election Systems & Software -- machines that are used in Allegheny County -- before they were certified by the federal Election Assistance Commission and the Pennsylvania Department of State.
Mr. Pollich said the audit showed that the software in the county's machines is an exact, true and unaltered version of the federally certified software held by SysTest and the state.
If even one machine had failed the audit, Mr. Pollich added, he would have recommended that the county immediately reload a new copy of the federally certified software.
Dr. Eckhardt and other VoteAllegheny members told the board of elections that the county should have used a bigger sample size and included the public in the audit process.
"I am pleased to hear that you all agree [software testing] is important, but what plans do you have for post-election analysis?" Collin Lynch, president of VoteAllegheny, asked the board. However admirable the county's efforts at software authentication are, he added, it should have chosen a statistically significant number of machines for its audit.
"Based on what [county and SysTest officials] said it doesn't sound like they used a random sample. And they didn't write down [their audit protocol] and didn't make it available to the public," Dr. Eckhardt said, adding: "We are not satisfied with what they did. Nobody can evaluate what the county did."
Tim Johnson, director of the Department of Administrative Services, which oversees the elections division, said the audit, which was done under tight security and video surveillance, was not open to observers because of the nature of the audit.
"We needed to have a controlled environment because we have never done this. We had to figure out what we were doing and how we would implement it," Mr. Johnson said, adding that his office will soon issue a report on the findings of the audit and make available the video of the actual four-hour test run on the 10 voting machines.
