'Crash tested' e-voting machines spread doubt on Super Tuesday
Six states at 'high risk'
In the Democratic and Republican primaries being held today, voters from 24 states will cast ballots for presidential candidates, making it the biggest "Super Tuesday" in US history.
But this election day comes with a much more dubious distinction: mistrust of the electronic equipment that will be used to tally many of the votes is higher than ever, computer and political experts say. Doubts about e-voting are no longer the esoteric stuff of geeks and conspiracy theorists. For perhaps the first time, they have become a mainstream obsession.
"In 2002 and 2003, I had trouble getting any credibility in the press or getting politicians to listen to me," says David Dill, a Stanford University computer science professor and a critic of electronic voting machines. "Now, the tide has definitely turned, and the momentum is against e-voting. By and large, there is a perception that it is problematic."
According to a report issued by two voting advocacy groups, six of the 24 states holding primaries today are at a high risk of miscounting votes because of machines that malfunction or are tampered with. Five other states are rated at medium risk, according to Common Cause and Verified Voting Foundation.
The states rated to carry the most risk - which include New York, New Jersey and Arkansas - are those that use electronic voting machines that don't produce a paper record that can be used in an audit or recount. Those found to be a medium risk use machines that provide a so-called voter-verified paper record but don't require audits that check for the accuracy of e-voting gear.
The widespread doubts about e-voting follow last year's release of reports prepared by elections officials in California and Ohio that found critical vulnerabilities in all the machines currently in use. Among the findings:
- An un-patched Windows 2000 server used by systems made by Premier Election Systems (formerly Diebold) left them open to a host of documented vulnerabilities that could allow it to be controlled by an attacker.
- An undisclosed account in the software made by Hart InterCivic could allow an attacker to gain unauthorized access to officials' election management database.
- Physical locks in Sequoia's Edge system could be bypassed by unfastening screws.
Even before the reports were issued, voters had grown wary of e-voting after some highly improbable election results were recorded in a 2006 race in Florida's Sarasota County. Machines supplied by a company called Election Systems & Software (ES&S) showed that Republican Vern Buchanan edged out Democrat Christine Jennings by just 369 votes in the race for the state's 13th Congressional district. More than 18,000 of the ballots cast recorded no vote in the race, an "undervote" rate that was about nine times higher than other races. (Jennings has contested the results in court.)
Each report or event "has planted a question mark, or raised a little red flag in people's minds about whether we can trust the machines, and whether they will live up to the promises made about them," says David Wagner, a computer science professor at the University of California at Berkeley who participated in a top-to-bottom study of voting machines commissioned by California Secretary of State Debra Bowen. "Four years ago, this only got attention from really obscure corners."
Elections officials in California, Ohio Florida and Colorado have all either scrapped touch-screen voting or placed tough new limits on their use. While critics of e-voting have generally applauded the moves, not everyone is happy.
"The efforts, especially in California and Colorado and Ohio, have been to cast doubts and aspersions on the electronic voting equipment by the very tests that were conducted," says Stephen Weir, the County Clerk of California's Contra Costa County and the head of the California Association of Clerks and Election Officials. "We feel that was grossly unfair and really designed with a conclusion in mind, and that was to decertify the voting systems."
In Weir's mind, California's study was akin to the crash tests car makers perform on new models before they are delivered to consumers. Laboratory scientists strap a dummy into a vehicle that's had its brakes disconnected and is sent hurtling into a brick wall.
"Yeah, the dummy is going to go through the windshield every time," he says. Just like the crash tests, he argues, the e-voting studies didn't take into account the real-world protections that are provided by things like security workers at polling places. "None of us had a chance to say: 'By the way, here's what it looks like when you have brakes, seat belts, et cetera,'" he complains.
It's a critique that's shared with some political science experts, including Henry Brady, another professor at UC Berkeley. While he remains concerned about electronic voting's susceptibility to equipment failure or tampering, he says those risks are being exaggerated at the exclusion of others.
One such risk few people pay attention to is the use of optical scanning machines that require ballots to be transported to a central office before being processed. That leaves them vulnerable to all kinds of tampering.
"We're so focused on the security issue that we've sometimes gotten rid of e-voting machines with paper trails... and replaced them with optical scan systems with a central count," he says.
Another overlooked problem, he says, is the confusing layout elections officials sometimes choose for paper or touch-screen ballots. Sarasota County's high undercount rate in the 2006 election was most likely the result of a ballot form that included two races on the same page, a mistake that could have been made using paper ballots.
Brady's suspicion may be correct. But because the ES&S machines used by Sarasota County didn't provide paper receipts showing how, or if, each person voted, it's hard to know for sure. And it's criticisms like these that are perhaps the most common refrain among e-voting opponents.
Son of Hanging Chad
When computers are the sole means used to register a vote, there's nothing tactile or otherwise to review later if anomalies are found. That's a deficiency that's largely not found with older methods of voting. Even during the hanging chad debacle of 2000, there were punch cards that could be inspected.
"These machines are an unnecessary risk," says Brian Chess, chief scientist at Fortify Software, a security company that supplied software that was used by officials in California, Ohio and Florida to analyze the source code of touch-screen machines. "They weren't developed with state-of-the-art security in mind or robustness in mind." (Fortify has offered to make its software available free of charge to voting officials throughout the country so they can independently analyze their systems.)
He says the experience companies like Diebold have gained in building highly secure automatic teller machines is of little value when designing e-voting machines. That's because the requirements for the two machines are vastly different. ATMs collect copious amounts of information about who is using the machine, exactly what was transacted and when. Touch-screen machines, by contrast, require that ballots be cast in secret.
In many respects, the move to e-voting is the result of the contested presidential election of 2000, which was settled only after a 5-4 vote along party lines by justices of the US Supreme Court. The controversy brought attention to the aging fleet of analog voting systems and affirmed voters' resolve for equipment that would be more dependable.
But for a growing number people, the resulting reliance on computers represents a step backward, not only because they are perceived as more vulnerable to malfunctions and tampering but also because there's no easy way to know if the results they report are accurate.
"With e-voting, there's a greater danger that we could have a problem on a much wider scale," says Avi Rubin, a professor at Johns Hopkins University and a longtime critic of touch-screen voting machines. "If we have an e-voting system and you get a result that appears perfectly believable and is wrong, you would have no way of knowing that. We don't know, and that right there is cause for concern." ®