Original URL: http://www.theregister.co.uk/2004/05/10/us_e_voting/
E-voting promises US election tragicomedy
New DRE (Direct Recording Electronic) voting machines came under fire in Washington this week, with expert witnesses giving contradictory testimony to the US Election Assistance Commission on the perennially entertaining and confusing topic of touch screen balloting.
The Commission is a tiny outfit charged with the very great task of recommending standards for election gear, protocols and procedures throughout myriad precincts, and encompassing a smorgasbord of balloting kit, of which five or six different models of touch screen system comprise but a fraction.
Add to this that the Chairman, DeForest Soaries, expressed delight at news that college students are volunteering as election workers, because, as he observed, "young people understand technology", and you can see that a monumental November cock-up is brewing.
Election officials, by and large, and with the notable exception of California Secretary of State Kevin Shelley, are much enamored of touch-screen, or DRE, systems, for several quite respectable reasons.
For one thing, they prevent overvoting: it's simply impossible to choose both 'yes' and 'no' to a referendum question, or to vote for a candidate and his opponent by mistake, a problem that surfaces regularly with paper and punch-card ballots.
They also reduce undervoting, as a summary screen is produced before a ballot is finally cast so that voters have a chance to observe that they've neglected to choose a president, say. Undervoting often occurs because one can't endure the sight of either candidate and can't, in good conscience, vote for either one; but more often than not, a confused voter has simply overlooked one or two races. Touch screens help ensure that a citizen has deliberately neglected to vote in a particular race or on a particular referendum question.
And they provide better accessibility for handicapped voters, reducing their dependence on poll workers, and give a much clearer indication of voter intent. It's either yes, no, or nothing; there are no hanging chads, pregnant chads, or stray marks and crossing out of previous choices. A DRE machine may reflect a voter's choice inaccurately, certainly. One may well vote for the wrong candidate, but one will do so clearly. Election officials love this.
Election fraud is a tradition stretching back to the dawn of civilization. There is no system, whether paper, mechanical, optical, or computerized, that can prevent it. We are clever little apes, and we will muck about with things. However, there are ways of making it evident that fraud or a malfuncion has occurred, and in this respect, touch screen devices are sorely lacking.
It's not so much that they're easier to scam than other systems; they're not. In fact, they're more difficult to scam than most. It's just that, as they're currently designed, it's difficult to detect tampering.
A box full of paper ballots can be sealed. If the seal is broken before the votes are tallied, it's painfully obvious that something has gone wrong. On the other hand, it takes some skill and planning to attack a DRE system; you don't just open a hatch and shove in a fistful of bogus ballots. But there are numerous points vulnerability; and while it may be more of a challenge to mount an attack, it's considerably easier to get away with one.
Comedian Jerry Seinfeld once said that the worst thing about being blind would be the inability to see if there were bugs in your food. This is the essence of the DRE problem: the devices offer a number of potential advantages, and if designed properly, could eliminate or mitigate a slew of serious problems, only the user has no way of knowing if they're crawling with bugs.
The voter verifiable paper record has become the clarion call of DRE skeptics. Their reasoning is simple: if a machine is suspect, the paper printout creates a separate medium for a presumably reliable recount. Only there are two problems: first, many election regulations specify that a recount must be performed in the same manner as the original election. Thus, if the machines are in use, such laws would require that the memory devices be read again, yielding the same, meaningless result ad infinitum.
Many devices have storage media as well as memory modules. But it is not known what would happen if the two should record different results, either due to an attack or a malfunction. Which is paramount? And if one should be empty, is the other an acceptable substitute without corroboration from a third source of data?
Second, in venues where paper printouts could legally be tallied in lieu of the electronic result, accuracy can only be assured if all voters review their recipts carefully, assuming they recall what they decided on fifty or a hundred questions moments earlier - a significant challenge. Furthermore, there are man-in-the-middle attacks that could potentially record the voter's input correctly on the paper record, but stealthily tweak the electronic results. And let's not forget that the paper output could be manipulated, though presumably, some sharp-eyed voter will notice this. If the two sets of results differ, the question then would be, which is paramount? The paper record is useful only if it is paramount, but as previously noted, there are numerous precincts where it probably won't be, and others where it will. Such uncertainties could entertain the courts for months - far past the deadline for election certification.
And the DRE machines are quite hackable at the moment. Johns Hopkins University computer science professor Aviel Rubin testified that "not only have the vendors not implemented the security safeguards that are possible, they have not even correctly implemented the ones that are easy."
Witness Neil McClure of vendor Hart Intercivic recommended that DRE manufacturers be required to implement the FIPS 140-2 crypto module standard. This would be "a great first step toward putting DREs on a path to becoming a trusted computing device," he reckoned.
No fan of the paper-trail panacea, McClure claimed that "irregularities can be traced to product quality issues," and advocated raising quality requirements and implementing national quality-management systems and testing requirements for all voting devices. Which of course will not happen, at least not between now and November.
Diebold Election Systems marketing director Mark Radke begged to differ with the curmudgeons and nay-sayers. His company's equipment is flawless, he implied - so flawless that it makes no sense to test whether or not it really is, because it has to be.
Asked to explain the basis for Diebold's claim that its kit is "eight times more accurate than paper balloting", Radke replied: "That would be based on such things as undervoting statistics and so on, against statistical fact, based on the information that we had for those elections."
"So when you use the word 'accuracy', you haven't really taken into account the possibility of tampering?" he was asked.
"Actually, no," he allowed. "We feel that our system is very secure, so that is not taken into consideration."
And there have been no reports of fraud or tampering, he added.
When questioned about uncertified software patches given to Georgia officials, Radke explained that these involved merely "a modification to the operating system, not to the tabulation software on our touch screen voting systems. It did not affect the tabulation process at all.
"We had a situation where, quite honestly, we had a few strains that we had difficulties on some of the units, and it was affected by the operating system; so since it did not affect the tabulation process at all, and did not affect that software, the operating system was modified. And, after those modifications were done, all the logic and accuracy testing was done, so every touch screen was tested before it was deployed."
It remains to be seen if such assurances will satisfy a technology-suspcious electorate when the inevitable tight-race disputes emerge, or if something far worse than Florida in 2000 will result. The public may not be in any position to judge how reliable DREs are in reality, but the constant example of Microsoft and its monthly worm debacles has persuaded many that computers can't be trusted.
Unfortunately, voter confidence in DRE currently depends on a Commission with little power and even less understanding, left to issue recommendations for election improvements based on conflicting testimony, and burdened by a September or thereabouts deadline for final certification of all equipment.
Ach, Chad, we hardly knew ye. ®
Thomas C Greene is the author of Computer Security for the Home and Small Office, a complete guide to system hardening, online anonymity, encryption, and data hygiene for Windows and Linux.
California preps e-voting ban bill
California decertifies Diebold bugware
Ireland to scrap e-voting plan
California set to reject Diebold e-voting machines
Judge OKs California e-voting
Pentagon cans Internet voting system
UK not ready for e-voting