From a statement:
To take over the voting machine, the computer scientists found a flaw in its software that could be exploited with return-oriented programming. But before they could find a flaw in the software, they had to reverse engineer the machine's software and its hardware-without the benefit of source code.
Essentially, return-oriented programming is a technique that uses pieces of existing system code to exploit the system. In this demonstration, the researchers successfully performed a buffer-overflow.
The team of scientists involved in the study included Shacham, as well as researchers from the University of Michigan and Princeton University. The hacked voting system was a Sequoia AVC Advantage electronic voting machine.
Shacham concluded that paper-based elections are the ay to go. I wouldn't go that far, but he did:
"Based on our understanding of security and computer technology, it looks like paper-based elections are the way to go. Probably the best approach would involve fast optical scanners reading paper ballots. These kinds of paper-based systems are amenable to statistical audits, which is something the election security research community is shifting to."
I'd settle for verifiable paper-based audit trail. Professor Edward Felten, a long-time observer of electronic voting systems also commented:
"This research shows that voting machines must be secure even against attacks that were not yet invented when the machines were designed and sold. Preventing not-yet-discovered attacks requires an extraordinary level of security engineering, or the use of safeguards such as voter-verified paper ballots," said Edward Felten, an author on the new study; Director of the Center for Information Technology Policy; and Professor of Computer Science and Public Affairs at Princeton University.
In February 2008, Felten demonstrated how he was able to access several electronic voting systems at multiple locations in New Jersey.