Using an earlier version of the source code that powers machines manufactured by Diebold Election Systems, the security experts--three from Johns Hopkins University and a colleague from Rice University--performed an audit and found numerous security holes.
"Our analysis shows that this voting system is far below even the most minimal security standards applicable in other contexts," said the researchers in a paper published Wednesday on the Internet, concluding that "as a society, we must carefully consider the risks inherent in electronic voting, as it places our very democracy at risk."
The criticisms echo a fundamental issue that many security researchers have raised with most current systems: There is no way to verify that a vote was correctly recorded and no permanent record is kept.
The issues also come as direct recording electronic (DRE) voting systems are taking off. In the 2002 election, 19.6 percent of the electorate could have cast an electronic vote, up from 7.9 percent in 1996, according to Election Data Services.
Diebold, the parent company of the election systems maker, assailed the researchers for using code that was out of date and never actually used in an election. Moreover, the company took issue with the report's focus on only the software and not the other security measures that the company uses.
"While respecting the report, we reserve judgment on the researcher's fundamental conclusions," the company said in a statement. "Our election systems products and services undergo a series of certification processes, which are conducted by federal, state and local officials, including logic and accuracy testing, and represent a sequence of security layers."
The researchers were only able to see the source code to Diebold's system because the company accidentally left the file on an Internet server in January. Manufacturers typically require that security experts sign a nondisclosure agreement before auditing the code.
Several issues became evident when the code was audited, said Avi Rubin, an associate professor of computer science at Johns Hopkins University and one of the authors of the paper.
For one, the manufacturer chose Windows CE as the operating system--a bad choice from a security standard, Rubin said. "Windows has a long history of new releases of patch just about every week," he said. "You can't run voting machines on Windows."
Moreover, the smart cards used by the system to limit a voter to a single vote could be duplicated. By bringing a stack of valid cards to the voting booth, a person could cast several votes.
Rubin warned that other systems are likely to have similar problems and of a similar magnitude.
Yet, not everyone is worried.
David Heller, the project manager for Maryland's voting system implementation, adopted a wait-and-see attitude. The state announced Monday that it had signed a $55.6 million deal with Diebold for the company to provide Maryland's voting systems.
"We are going to analyze the report and make intelligent decisions," he said. "We have been using the systems now for a year and a half. We've had great success in dealing with them."
Heller explained that the aspects of the system not analyzed by the researchers do make a difference. On election day for example, human election workers would count the number of votes cast at each terminal and retain receipts that would tie people to a specific machine (but not to their actual vote). If the voting machine's tally doesn't match the operator's count, then the votes on that machine would be thrown out and those voters allowed to recast their ballots.
"If there is a failure or a compromise of one unit, we know who voted on that unit," Heller said. "We don't know how they voted, but we know who voted."
Still, the researchers argued that the software needs to be secure to prevent abuse of the election process.
Tadayoshi Kohno, a researcher at Information Security Institute at Johns Hopkins University and an author of the paper, said the secrecy that surrounds the creation and certification of electronic voting machines almost ensures that the systems will not be programmed securely.
"Independent security researchers need to verify and look at the source code of electronic voting systems," he said. "There are no easy fixes. I think that as a society, we are moving too fast toward electronic voting and we need to rethink things more thoroughly."