Travis County Clerk Dana DeBeauvoir has spent more than a decade working with researchers and computer security experts to design a voting machine that’s more secure and reliable.
This massive undertaking resulted in the Secure, Transparent, Auditable, and Reliable Voting System, or STAR-Vote. But getting manufacturers to build it has been a challenge.
DeBeauvoir says it all started around 2005.
“I got a little irritated because many of the folks in the computer security community were criticizing election administrators for what they saw as the inadequacies in DREs," she says, referring to direct-recording electronic voting machines.
Security experts have a lot of issues with electronic voting. Their logic is pretty simple: Computers can be hacked; paper can’t.
But after serious problems with paper ballot machines during the 2000 presidential election, voting administrators ditched paper ballots and started using DREs.
When Houston first floated the idea of switching to DREs in 2001, it caught Dan Wallach's attention. He urged city leaders not to ditch paper ballots.
“My message then was: These are just computers,” says Wallach, a professor in the department of computer science at Rice University, “and computers are hackable.”
Houston switched to DREs anyway, and so did Travis County – along with much of the state and a lot of the country.
Around that time, DeBeauvoir and Wallach started having run-ins at public hearings.
“I would say, ‘This is a bad idea,’" Wallach says. "And she’d say, ‘No it’s great.’”
The tension between them represents a significant divide between people in the voting world.
Some experts, mostly those concerned with security, swear by paper ballots. Others, like DeBeauvoir, prefer electronic voting. She says paper ballots aren’t as helpful for voters with disabilities.
“You have to comply with the Americans with Disabilities Act in whatever voting system you have,” she says. “And the only voting systems that assist voters with disabilities are electronic voting systems.”
DeBeauvoir says you can have two different systems for people who are disabled and people who are not, but that’s not ideal. Paper ballots also make compliance with another law harder, too, she says.
“Texas is what we call an ‘intent-of-the-voter’ state,” she says. “So, if you are going to use a hand-marked ballot, you have to run every single piece of paper before a set of human eyes before it goes to a machine to be read.”
That makes the process slower and more expensive, she says.
Years later, Wallach and DeBeauvoir would cross paths again when DeBeauvoir was invited to speak at a voting technologists conference.
“I challenged the group," she says, "and I told them, ‘Look, if you want a better voting system, then work with us.’”
She told the crowd of people – many of whom were her critics – that she would help them develop a voting system that was more secure but that could also work in the field.
“We were all like, ‘What?’" Wallach says. “And what she said was, ‘Well, our machines are now a decade old, they are starting to reach the end of their service lifetime, we will have to replace them, and we don’t like anything that’s on the market.’”
After the conference, DeBeauvoir made good on this and called Wallach out of the blue.
“And she says, ‘Dan I want your help. Let’s do this,’” Wallach says. “I said, ‘Wow, sure. Can I invite my friends?’”
His friends – other researchers and computer security experts – all agreed to help DeBeauvoir out.
“And next thing I know, we have a meeting in the Travis County Clerk’s office with her elections people and a whole collection of different security and usability and statistics people,” Wallach says.
The team would spend years working on the project.
“We worked for 12 years together," DeBeauvoir says, "a group of some brilliant people, to put together what ultimately became STAR-Vote.”
What the group came up with basically combined paper voting and electronic voting in one system. DeBeauvoir says it’s an electronic voting machine that creates a paper trail.
“This is the matching paper document for what the voter created in an electronic record of their actual ballot,” she says.
The paper record would be linked to the electronic vote, so auditors could see if there were any discrepancies.
“You don’t have to necessarily have to refer only to a recount to confirm the total of the elections,” DeBeauvoir says.
And, she says, those audits could be done by a third-party, which makes the design even more transparent.
Another key feature of STAR-Vote is that voters would have verification that their votes were cast the way they intended.
Josh Benaloh, a senior cryptographer with Microsoft Research, says this aspect is important to him, because it means voters wouldn’t have to blindly trust machines or election officials.
“People can check for themselves,” he says. "And this is a very powerful property. [It can] even be said to be a democratization of the electoral process that I’d very much like to see in Austin and all over country and all over the world.”
Philip Stark, the associate dean of the Division of Mathematical and Physical Sciences at University of California-Berkeley, helped with the design. He says this is why it was important to bring computer scientists and researchers to the table.
“The electronic side of STAR-Vote is pretty complicated and pretty cool – using cryptographic methods to allow voters to confirm that their vote ultimately was recorded and ultimately included in the total correctly,” he says. “So, I call this a ‘belt and suspenders’ model for election integrity.”
Combined with an easier design of what voters actually interact with on these machines, STAR-Vote became something that election officials like DeBeauvoir and Wallach could agree on.
“I wish that happened every day, but it doesn’t,” Wallach says.
The process alone was also something that had never been done before.
“Dana had the courage to go out and basically stick her neck out and say, ‘OK, none of the vendors out there are providing anything close to what my constituents need. I am going to reach out to researchers and see if we can put something together and see if we can get it built,’” Benaloh says.
But that last part, actually getting STAR-Vote built, is where this ambitious project met some frustrating realities.
Once DeBeauvoir and her team finalized their design for STAR-Vote, they took the next step and tried to find a company that would build the machine for them. DeBeauvoir reached out to manufacturers and opened up the bidding process.
“We received 12 different bids,” she says. “Twelve proposals from the biggest names and the best people in the manufacturing communities for voting systems.”
The STAR-Vote team broke up the design so companies could bid on different parts of it.
But the team ran into a problem: No single company wanted to build an important part of STAR-Vote: its open-source system, which meant the software would be publicly available and modifiable.
“There was not a complete build in all of those proposals,” DeBeauvoir says. “So, it was very disappointing."
One of the innovative ideas behind STAR-Vote is that counties could buy hardware off the shelf – which election administrators are more likely to afford – and the technology could be run on that hardware.
“I am not surprised that none of the existing manufacturers of voting systems wanted to bid on it, because it’s basically a radically different model that competes with theirs,” Stark says.
Stark was hoping someone outside the existing voting machine marketplace would be interested, though.
“I did hope that some entrepreneurial people would get together and say, ‘This is an interesting thing. Let’s change the world and build a better voting system,’” he says. “And unfortunately, that didn’t happen.”
A company could still look at STAR-Vote – it’s pretty easy to find the design online – and build its own.
Wallach says the idea doesn't have to wait for manufacturers to shoulder the front-end costs.
“If we found the right philanthropist who wanted to spend, let’s call it $10 million, to fund the development and design and to get it to a point where it is ready to rock and certified and tested," he says. "And then it’s just a thing that counties nationwide could choose, among other things."
And instead of a single county trying to get a more reliable and secure voting system in place, Wallach says, a whole state could do it.
“The State of Texas could spend the money once to develop the software, and it would be something that counties could use,” he says. “At that point you have the economies of scale where it becomes financially effective.”
Stark says he thinks STAR-Vote could be built by the federal government.
“I mean, I really think that our country should own its own voting systems,” he says. “I don’t think we should be relying on the private sector for the technology that we use as the backbone of our democracy.”
Ultimately, DeBeauvoir says, there is a path forward for a version of STAR-Vote in Travis County. She says voters will see new machines before the next presidential election.
The Travis County Clerk’s office has a new request-for-proposal that includes most of the technical specifications in STAR-Vote’s initial bid, and she says it will be a big improvement.
“We are not going to get STAR-Vote, but we are going to get something really close to it. And I am really happy the marketplace has changed,” she says.
The marketplace could soon have an electronic voting system with a paper trail that is encrypted and easy to audit.
DeBeauvoir says she’s glad this whole STAR-Vote journey happened, that a positive outcome came out of the years of frustration.
“I took that anger and converted it into a challenge,” she says. “And that has turned out to be one of the most wonderful working experiences I could have ever hoped for.”
DeBeauvoir says STAR-Vote also inspired some longtime changes to federal standards for voting machines.
The U.S. Election Assistance Commission is updating its Voluntary Voting System Guidelines, which could set new standards for future voting machines.
Many of the innovations from the STAR-Vote team are set to be included in those standards.
This post has been updated.