MIT E-Voting Project To Analyze Experience Of Voters in Election

Originally published in The Tech on September 30, 2008

As the election approaches, a question lingers with increasing urgency on the minds of concerned citizens: to e-vote or not to e-vote?

The question, of course, cannot be defined so simplistically, as electronic voting refers to several types of voting that include electronic means of both casting and counting the vote. These systems both present pluses and minuses, and have been the focus of a public backlash amid growing security problems.

Following the Florida fiasco of 2000, the nation’s recent rush to embrace e-voting has been hampered by controversy over issues of security, reliability, privacy, and cost.

The release earlier this month of a short video by a group of scientists at the University of California Santa Barbara showing how a single person can hack an election on a Sequoia touch-screen voting system undetectably is the latest in a series of such demonstrations over the past two years. The hack demonstration also shows how the so-called “security-seals” placed on such machines after they have been programmed for an election can be easily defeated without detection.

These enormous flaws reinforce the insecurities surrounding the use of electronic voting technology and the need for better technology and legislation.

Amid growing evidence of security flaws and pressure to stop the use of similar systems in American elections, researchers at the Caltech-MIT Voting Technology Project are hard at work analyzing the security and impact of e-voting systems.

The Caltech-MIT Project

“It’s a general program that studies a lot of issues, including e-voting. One of the issues we are working on is trying to figure out ways that voters can make sure their vote has been counted, without receiving a paper receipt,” said Charles H. Stewart III, head of MIT’s political science department and a member of the VTP faculty. “You don’t want to carry a piece of paper around,” he said in a telephone interview last week.

Established by the California Institute of Technology President David Baltimore and MIT President Charles M. Vest in December 2000 to respond to the need for academic guidance in the wake of the 2000 U.S. Presidential Election, the Caltech-MIT VTP seeks to improve voting in the United States through the use of the latest technologies. Its tasks include evaluating the reliability and administrative practices of existing voting systems, establishing guidelines for their reliability and performance, and proposing standards for the design of new voting technologies.

Judging by the titles of the posts on the initiative’s blog, they still have a lot of work to do. “Plan for the worst,” the Election Assistance Commission urges in its Election Management Guidelines, and “Voter Registration Problems Reported in Michigan,” the blog quotes the Detroit Free Press as saying.

With regards to the upcoming election, MIT is trying to collect as much data as possible through two data-related projects of the Caltech-MIT initiative.

The first one is simply collecting the Election Day returns from the states, so that one can see the difference electronic voting machines made, Stewart said. “All states report information about the machines they use and the problems they encountered.”

The other big data project is the development of the first large-scale public opinion poll of American voters to ask them about their experience with voting on Election Day. “We will sample over 10,000 voters, drawing samples in each state of the country. So, we will be able to gauge not only how well the election went nationally, we will also be able to tell whether particular states offered special problems, or went especially well,” Stewart said.

“Another big project we are working on is trying to understand empirically all the things that are related to voting machines,” Stewart said.

“One thing we are trying to look at is voting machines not recording the votes people have made. As what happened in Florida in 2004 when the votes had to be re-counted, some machines are more prone to break down, more prone to confuse voters. That’s what we study,” he said. “It’s an issue we call ‘lost ballots’.”

Auditing the election — a tricky task

As for whether MIT will try to audit this election, the MIT project is not an auditing project per se, Stewart said. “Some of us on the Caltech-MIT Project will be doing analyses after the election to try and understand the role that voting technologies played in the results of the election, which will have some characteristics of an audit,” he said.

“It’s a term that’s being thrown around a lot,” he added. “What my colleagues are going to do is research nationwide to see if the machines that are set up nationwide are prone to break down. That’s more a statistical task,” he added.

Auditing can at times be theoretical, and at times applied, Stewart said. “The amount of data that is produced is enormous.” Thus, he continued, “it’s physically impossible for a human being to take a look at every single paper ballot,” which is what people usually refer to when they talk about “auditing ballots.”

Ronald L. Rivest, professor of Electrical Engineering and Computer Science and a member of the VTP faculty, is doing research on auditing, elections and cryptography — a technology some voting systems use to ensure that votes were cast correctly, detect possible fraud or malfunction, and provide a means to audit the original machine.

Rivest, who is also a member of the technical advisory board of the Election Commission, which regulates voting, is currently working on a system that sample precincts.

In an interview, he stressed that MIT is not doing research on paperless voting systems.

“My own research does not involve paperless voting systems, although paper can be used in a variety of ways to improve a voting system. A purely electronic voting system would typically not be ‘software independent,’ and so you would have to trust that the software is correct and unmodified in order to have confidence that the election result is correct,” he said.

“By and large, paperless voting systems would be ‘software dependent,’ so that an undetected error or change in the software could cause an undetectable change in the election outcome. This would be very undesirable,” he said.

No MIT voting machines

As to whether MIT will one day make its own voting machines, Rivest said, “we do research on voting systems, and may produce prototypes. We don’t produce them for others to use, a vendor would do that.”

Stewart said MIT is not working on building electronic machines, but is studying the impact of electronic voting machines.

“That MIT might be working on its own machines is an idea that dates back to 2001,” Stewart said, “but we are less interested in making our own machines than in providing more technical information and fundamentals in science and engineering.”

As examples of MIT research closely tied to e-voting machines, he cited former Caltech-MIT VTP co-director Ted Selker’s work on electronic interfaces, protocols for securing votes electronically, and on audio systems that could read back votes cast on electronic voting machines. One application of the latter is for blind and visually-impaired people to have a way of having their vote read back to them so they can ensure it has been recorded correctly.

Jonathan Goler ’04, who designed and developed the Caltech-MIT VTP as a UROP, said in an interview earlier this month, “MIT will never produce its own e-voting machine. MIT will probably continue to release concepts for improving the electronic voting process, such as the Low Error Voting Interface and Secure Architecture for Voting Electronically.”