Princeton University computer scientists issue report and video showing how it’s done
By: Hilary Parker
In a presidential election between George Washington and Benedict Arnold, who would win?
It depends on whether hackers broke into the electronic voting machine, according to a paper and video released Wednesday by Princeton University computer scientists on the Web site for the university’s new Center for Information Technology Policy, itpolicy.princeton.edu/voting.
A 10-minute video excerpt on the Web site demonstrates how vote counts on electronic machines can be changed by software that is quickly installed and undetectable thereby securing a win in a staged election for Benedict Arnold in less than 60 seconds, though Washington actually earned more votes.
Center for Information Technology Policy Director Edward Felten and two graduate students, Ariel Feldman and Alex Halderman, created the software and demonstration using a Diebold AccuVote-TS unit they acquired in May.
"We found that the machine is vulnerable to a number of extremely serious attacks that undermine the accuracy and credibility of the vote counts it produces," they wrote.
The concern should extend to other electronic voting machines as well, Professor Felten noted, pointing out that a rigorous study of other similar voting machines would likely have yielded similar results.
The paper came just one day after U.S. Rep. Rush Holt (D-12) spoke out on reported problems with electronic voting machines in the Maryland and District of Columbia primary elections.
Long an advocate for national audit standards for electronic voting machines, Rep. Holt’s Voter Confidence and Increased Accessibility Act of 2005 (H.R. 550) is due to come before of the House Administration Committee on Sept. 28.
"This report shows that stealing electronic votes takes less time than making five-minute rice," Rep. Holt said in a written statement. "They’ve done exactly what all of us have feared possible: successfully hacked a voting machine and changed the outcome of an election. We should count ourselves lucky that it was a mock election but only if we act today to make it impossible before Nov. 7, when real votes and real elections will be on the line."
Two of the main provisions of the act the requirement of a voter-verified paper trail and a random audit of results match key recommendations made by the researchers in the paper.
The study may have pleased Rep. Holt, but it did not earn Diebold Election Systems President Dave Byrd’s vote of confidence.
"By any standard academic or common sense the study is unrealistic and inaccurate," he said in a written statement. The AccuVote-TS unit that the university researchers studied featured out-of-date software, he said, and electronic and physical security measures were ignored throughout the demonstration.
Mr. Byrd noted that voting machines are never networked and the newest models all feature state-of-the-art security, including advanced data encryption technology.
But viruses don’t need networks to spread, Professor Felten said, and can easily spread via memory cards that are transferred between machines in normal pre- and post-election activity.
The argument about physical barriers to election tampering is no stronger, Professor Felten said. After all, he noted, in the demonstration video that accompanies the paper, it took him and his colleagues just seven seconds to pick the lock, and keys to the locks are readily available anyway. As for the screws that secure the information within the machines, Professor Felten said it just takes a screwdriver.
"Screws are not an effective security measure," he said. "If the security of the election is relying on the inability of a bad guy to unscrew a dozen screws, we’re in deep, deep trouble."