E-voting vent: You can't tell if it worked

Paul Andrews   Seattle Times  20 September 2004

By most indicators, Frank Love Elementary School in Bothell was just another suburban voting location last week.

Makeshift "Polling Place" signs and storybook paw prints in blue paint marked the path to the school library, which housed six voting booths. The room was quiet and orderly, and poll personnel were helpful and personable.

Still, Fred McLain had to figuratively hold his nose when he cast his ballot Tuesday morning. McLain voted using a computer equipped with a touch screen, but it lacked any method for him to confirm that his vote had been recorded properly and counted correctly.

For most e-voters in Snohomish County — the first in Washington state to implement touch screens two years ago — the lack of an auditable paper trail was no big deal. E-voters I talked to at Frank Love said they liked the machines for their speed and convenience. Most said they trusted that their vote would be counted correctly, adding that any system, including paper ballots, is potentially flawed.

McLain, a software professional who served as lead programmer for an e-voting system devised by a reformist group called the Open Voting Consortium, isn't so sanguine. Lots of mistakes, unintentional or otherwise, can happen with e-voting, he says. Yet with a few simple steps, the process could be made far more accountable.

To start with, he says, a lot of concern could be cleared up if commercial touch-screen providers would permit their software code to be viewed by independent experts.

For the most part, they haven't. Source code — the brains inside the metaphorical "black box" — is held secret by voting-machine vendors. The Open Voting Consortium, by contrast, uses free, visible, open-source code for its system.

At Frank Love School, voters were issued a "smart card" to in a slot at a terminal. That gave them an opening screen, where they were asked to declare their party preference. Once they chose an affiliation, subsequent screens listed only that party's candidates. 
 

McLain had no problem using the system and entering his votes. One thing did stop him: The final screen, presenting his ballot in overview, highlighted a missed entry in red. It was a write-in-only category, and McLain hadn't entered a name.

McLain's 18-year-old son, who would be voting after school, is colorblind. "How's he supposed to recognize the highlighting?" McLain asked. (County Auditor Bob Terwilliger acknowledged the colorblind issue and said it would be addressed "at some point" once more pressing issues were resolved.)

E-voting critics like McLain are most troubled by what happens after the ballot is cast.

Under the Open Voting Consortium process, voters would be issued a paper version of their ballot — as cast — for review.

The ballot would contain a bar-code record of its contents that would be tabulated electronically after the voter approved it. Voters have the reassurance that, in case of a recount, their printed ballot would reflect their vote accurately. They also know a paper trail exists in case a recount is required.

As it stands, each Snohomish County terminal compiles a printout of its contents, noted Don Freeman, depot judge at Frank Love Elementary. But the paper is merely a hard-copy backup of the machine's counts and not a ballot-by-ballot record of the vote. Voters have no physical proof their ballot was recorded and counted accurately.

In response to such concerns, Terwilliger notes that the county ran a "parallel monitoring" process using representatives from each party and test ballots on isolated machines throughout the day. The process was designed to catch any software worms or Trojan horses that, if contained in the code, could have been scheduled to activate during the day and change the vote counts.

A review by party and elections officials of the machine results after polls closed showed no discrepancies, Terwilliger said.

The biggest problem the county encountered was the disabling of 65 of 860 e-voting machines because of a software crash and jamming of the smart cards. Although affected voters managed to use alternate machines and no previously tallied votes were lost, "we can't afford to have a repeat" of the breakdown during the November election, Terwilliger said.

He has asked Sequoia Voting Systems, the machine vendor, to investigate the problem.

McLain's confidence is further undermined by the absence of federal certification of software devised for the new primary voting system (King, Pierce and three other counties also lacked certification due to the change). Because of the relatively late primary-ballot change, Terwilliger said, only provisional certification through the Secretary of State's Office could be obtained. Federal certification will be done, but not till next year, Terwilliger added.

Given nationwide concerns over e-voting and the polarized nature of November's presidential election, McLain argues that every effort should be exhausted to close loopholes in the e-voting process.

"Even though this [e-voting] has been going on in Washington state since 2002, the standard response of elections officials is that there isn't enough time to correct anything," McLain said. The excuse "is getting a little hollow."