Harris is convinced that wholesale election fraud has been going on in the United States, hidden behind the technological complexity of the new electronic voting systems that are being adopted across the nation. For example, she uncovered the fact that a senator from Nebraska who won surprising upset victories in his maiden try for office in 1996 had been, until shortly before, the chairman of American Information Systems (now called Election Systems & Software), the company that made the voting machines used to elect him. These and other charges are laid out in her book, Black-Box Voting: Ballot Tampering in the 21st Century, and on her website, www.blackboxvoting.org.

Harris's claims have the aura of conspiracy theory, and one can't help noticing that most of her targets are Republicans. Still, that doesn't mean she isn't right. Is there really a problem here?

The US Presidential election of 2004 has come and gone. The election, in which for the first time 30% of the votes were cast on direct-recording electronic (DRE) voting machines, seems to have gone relatively smoothly, all things considered. But does that mean that we have nothing to worry about? Or might the machines have concealed widespread errors or even outright fraud?

As in 2000, the election yielded a victory for Bush, and as in 2000, the Democratic pundits are now wringing their hands and searching for an explanation for the disaster that has struck their party. This time, poorly designed ballots and hanging chads do not seem to have played a major role. Instead, the new culprits are said to be the DRE voting machines. Even before the election, criticism of these machines was widespread. As a measure of the amount of distrust that has been in circulation, consider the brouhaha that erupted over a statement made in 2003 by the CEO of Diebold, one of the major DRE makers. He also happens to be active in the Republican organization in Ohio. In an invitation he mailed out for a fund-raising dinner, he said that he was "committed to helping Ohio deliver its electoral votes to the president next year." This remark was probably nothing more than political boilerplate, but it was seized upon as evidence that Diebold Election Systems was not to be trusted to provide the voting machines for the state of Ohio. It should of course be noted that the loudest voice raised in protest was from Sequoia Voting Systems, a rival DRE vendor that had lost out to Diebold in bidding to provide voting machines for the state.

In fact, it does not seem likely that the makers of DRE voting machines deliberately skewed the election in favor of Bush, and certainly no one has provided any solid evidence that they did. The pundits may just have to accept the fact that the country voted him in. Nonetheless, the questions that have been raised about DRE machines deserve investigation.

Electronic voting systems are not a brand-new phenomenon. Machines to scan and tally paper ballot forms electronically have been in use for decades. DRE machines began to be used in the 1990s. They have been adopted widely in a number of countries, including, perhaps surprisingly, India and Brazil. But they remained a small segment of the voting-machine market in the US until the 2000 election, in which the outcome hinged on problems with "old-fashioned" technology. Even those who were pleased by the outcome didn't want to see a repetition of this circus, and the result was the Help America Vote Act (HAVA) of 2002. Suddenly a huge pile of cash was available for states to buy new and improved voting systems. Overnight, electronic voting was transformed into a huge boondoggle, with DRE vendors competing fiercely to get the biggest slice of the US$3.9 billion pie. It would not be surprising if some corners were cut and quality-control steps skipped along the way to getting their machines to market in time for the election.

Aviel Rubin is a computer scientist at Johns Hopkins who specializes in computer security and applied cryptography. In June 2003, following a tip from Bev Harris, he obtained a copy of the software for a complete voting system that had been left on an unsecured ftp site operated by Diebold. Although it was never proved that this software was actually used on any systems Diebold sold, it seems reasonable to assume that it was at least closely related to production code. Rubin and his colleagues found numerous problems with the implementation of security features, which left the system vulnerable to numerous modes of attack. Their most serious criticism of the software, however, was simply the lack of professionalism or good software engineering practice that it revealed.

All this is not to say that electronic voting is an irremediably flawed concept. If implemented properly, it can be secure, and it may be easier and less confusing to use than some of the older methods of voting. For instance, typical systems use touch screens so that voters do not need facility with a mouse and keyboard. Voters who are blind can be provided with headphones for aural cues.

Much has been made of the need for a "paper trail" for purposes of reassuring the voter that his or her vote has been properly recorded, and to provide a means for a recount in case the vote totals are questioned. It is rather ironic that the call for paper ballot printouts has been coming mainly from the computer experts, who one would expect to be gung-ho for all new technological gizmos. But these experts agree that the state of the art does not yet allow a totally electronic audit trail to be securely created, leaving paper as the simplest and most reliable fix for now. Fortunately, it is not difficult to add this capability to most DRE machines on the market.

Vote-rigging is another potential problem, and in contrast to traditional voting methods, a compromised computer program could affect results on a wide scale. One suggestion for helping to increase trust that the voting software does not contain hidden dirty tricks is to make the source code available for all to read. This is actually the practice in Brazil, where the voting machines were designed by a government agency, but in the US the DRE makers protect their source code as a trade secret, so opening the source code would require a drastic change in the way that the voting systems are developed and sold.

Of course, older voting systems have problems of their own. Paper ballots can be confusingly laid out and cannot have the smarts to check for invalidating errors before the vote is cast. Lever machines are somewhat better in both respects, but the companies that built them went out of business decades ago, and the aging machines need careful maintenance. These also lack an audit trail for a recount. Given the problems with the older technologies, DRE machines could be an improvement, making voting easier and more accessible than before.

In any case, e-voting is here to stay, for better or for worse. In the final analysis, it appears that electronic voting systems are not really all that much more vulnerable to failures or vote-rigging than traditional systems. The real question is, how can we tell whether the machines worked correctly and reported honest results? What is most important is transparency of process and cooperation among all interested parties in implementing the voting systems that are used, to ensure there will be the necessary level of trust in the validity of the outcome.

Associate Professor Bob Moniot teaches computing at Fordham University in New York. Email moniot(at)fordham.edu