Princeton professors examine electronic voting

By Alexander Maugeri

The Daily Princetonian (Princeton U.)
01/29/2004

(U-WIRE) PRINCETON, N.J. A little over half of Americans vote for president, less than a third for members of Congress. Most say it's because their vote doesn't matter. But what if votes didn't matter. What if votes were being systematically eradicated and discounted?

As electronic voting machines become more popular around the country, Princeton University is becoming something of a focal point for a critical discussion of the new technology. Two Princeton professors and New Jersey congressman are focusing in on the issue.

Congressman Rush Holt, Democratic representative for central New Jersey including Princeton, is concerned that fear of "hanging chads and butterfly ballots" is prompting too many communities to buy potentially unreliable electronic voting machines. He believes current legislation the Help America Vote Act of 2002 is inadequate and is proposing a new bill in the House.

HAVA requires that all states reform the voting processes, but beyond that its statutes are vague and have led to varied interpretation.

Diebold, Inc. is one of the largest distributors of electronic voting machines, with over 33,000 terminals nationwide. There are no proven cases of fraud or miscounting with Diebold machines as yet. However, this has not deterred critics.

These systems provide no "barometer for judging accuracy," says computer science professor Edward Felten. He contends that since all tallying takes place inside the system, voters have no way of knowing if their vote was truly registered.

"A programmer could put malicious code in the software, or there could be a bug," Felten said, explaining how the software itself is the greatest vulnerability.

On July 30, Diebold agreed. The company posted on its homepage that "a combination of malevolent insiders and unscrupulous voters could tamper with [election] results." But company spokespeople say any machines would be susceptible to that level of fraud. Therefore, they say, their technology can not be expected to guard against it.

Diebold has used copyright laws to quash internal memos and e-mails admitting to security flaws and refuses to make its voting machine software code available for independent inspection.

Last year the Johns Hopkins University computer science department concluded that "common voters, without any special privileges can cast unlimited votes without being detected by any mechanism in the voting terminal." Felten worries that without public scrutiny, under- or overvoting could go unnoticed.

Such was the case in a Fairfax, Va., school board election. Some voters noticed "when they pushed the button for a given candidate an X would appear over the candidate's name and then later disappear," Felten said. After testing, it was found that about 10 percent of votes were being invisibly ped.

Most e-voting critics say there is a need for a post-election audit. Holt proposed a bill in Congress, HR 2239, that would require that one half of one percent of all votes always be audited recounted for irregularities.

Electronic voting machine manufacturers generally agree that this would be good practice, however the question of what to recount is a major point of contention. Diebold and others say their terminals have a built-in audit capability and can verify a percentage of votes electronically.

Holt does not trust "some company" to regulate the safety of its own product. He proposes a paper ballot be printed at the terminal, and that each voter check its results against the computer screen. Diebold said in an Internet rebuttal to the Hopkins study, "this system would essentially reduce an electronic system to a paper system, which has risks of its own."

Critics in the media have said that the paper trail remedy will increase the public's thirst for a recount and bog down the electoral system. Human error in counting paper ballots could produce a recount more inaccurate than the first. Also, a paper recount would be more susceptible to traditional forms of fraud, such as ballot stuffing. Felten and Holt agreed that a paper count could not feasibly be the official tally in an electronic system.

Wilson School professor and New York Times columnist Paul Krugman has written in his columns that he is concerned with "extremely sloppy security" at Diebold, like the problems found by the John Hopkins researchers, and possible attempts "to cover up product defects" such as Diebold's attempt to prevent critics from posting its internal memos.

Krugman refused to comment, his assistant said, because he intends to author another New York Times editorial on the subject.

And, according to a December Associated Press story, Diebold's staff might include characters willing to engage in malicious actions. Jeffrey Dean, a chief programmer for the company, has spent time in a Washington, D.C., jail for embezzlement and tampering with computer files.

Software is not the only soft-point in electronic voting. Tampering in the election booth is worrisome, said Felten.

"It's far easier for an election worker, or voter for that matter, to modify a computer than one of those large mechanical voting machines."

Some machines just make phone calls to tally the votes without encoding the data, and in rare scenarios totals are sent using wireless network technology, which is more easily eavesped. Felten implied that these practices should be discontinued, and the latter would be prohibited under HR 2239.

The bill currently has about 100 cosponsors, but Holt said he doesn't believe it will pass in time for the 2004 presidential race. Holt says he's optimistic about the future of electronic voting. He believes with a verifiable paper trail the machines can be made to be safe and accurate. And, they have already been proven to be more handicap accessible and efficient than lever machines, a popular technology since the 1920s.

While singing the praises of the resolution, Felten was careful to emphasize that random audits will not ensure that each election is accurate, because small errors are inevitable. Instead, they would test the reliability rating for electronic voting machines in general as well as catch "large systematic fraud."

Reports are now emerging that the 2002 midterm elections may have been riddled with irregularities. If true, this could mean potential problems for 2004 since an additional 20 to 25 percent of voters are expected to participate. Critics of e-voting predict the year 2000 situation in Florida will reproduce itself in various other states. Yet, proponents are saying the upcoming presidential bid might be the most accurate election in U.S. history.