article in the Economist

["Margit Johansson" <margitjo@xxxxxxxxx>]

http://www.economist.com/opinion/displaystory.cfm?story_id=8028608

Crash and re-boot

Oct 12th 2006 (From The Economist print edition)

The wrong kind of voting machine could bring chaos to the mid-term elections

THE polls go up, the polls go down and there are still more than three
weeks to go: time for any amount of sleaze or terror to influence the
voters. But it is quite possible that America's mid-term elections on
November 7th will produce a close result, not just in the House of
Representatives, where it has long been predicted, but in the Senate
too. At which point things could get fraught.

The problem is voting machines. Not the ones with hole-punches and
their chads, hanging, swinging and dimpled. Since the debacle of 2000
in Florida federal money to the tune of several billion dollars has
been lavished on replacing them. Unfortunately, many have been
replaced with new ones that may be even worse. In a close election the
prospect of just a handful of the 435 House seats or one or two of the
33 Senate seats at stake being furiously challenged in court is all
too plausible. Like the presidency in 2000, the colour of Congress
could have to be decided by lawyers.

How could this have happened? Mainly because lots of states and
counties went for touch-screen devices, very like ATMs, instead of a
much better alternative, optical scanners that count votes marked by
hand on paper ballots, rather like lottery forms or multiple-choice
exam papers. The good thing about scanners is that the original ballot
is by definition available for re-counting. With touch-screens, it
isn't. Fortunately, more than half of America's 3,000-odd counties
have opted for scanners. But about a third have chosen the
touch-screens.

A thermal printer to produce a record is available as an added extra
on some touch-screen machines. But when these were tried out, as they
have been across the country in primary elections in the past few
months, the results were not encouraging. If the paper is put in
wrongly, the printer does not print at all. Even when it was the right
way round, there were many cases of the paper jamming, tearing or
producing unreadable results. The fact that most election officials
are unpaid volunteers, very often elderly and with little or no
training, also caused difficulties. At any rate, with a touch-screen
system the paper-trail is produced by the machine, and so is only as
good as the machine. And the machines, it also turns out, may be
vulnerable to tampering.

In September three scientists at Princeton University got hold of the
most popular touch-screen model and took it and its software to bits.
They found serious flaws allowing a competent hacker to infect the
machine with a program to transfer votes from one candidate to
another. Such a change could be undetectable without a recount
(assuming one were possible), and the program could be introduced into
the machine far in advance by anyone having access to the machine's
memory-card reader for as little as a minute. The readers are
protected by a lock, but the lock is a standard one, and keys can be
bought on the internet: besides, the keys circulate among election
officials. And the researchers found that their program could be
spread from machine to machine via the memory-cards. Voting-machine
companies make things worse by keeping their software secret: were it
published, security experts would be able to assess it and recommend
fixes.

Pray for a clear result—and then buy some decent kit

Many people will see this as a conspiracy too far, and perhaps it is.
But another problem is that the machines used at polling stations have
a tendency to break down. Avi Rubin, a computer science professor at
Johns Hopkins University who is also an election volunteer, has
blogged scarily about his experiences at the primaries in Maryland.
(In his case, most of the problems occurred not with the voting
machines but with other terminals, designed to ensure that people are
entitled to vote and have not voted twice.) Elsewhere, there have been
horror stories of votes failing to register or upload, of memory-cards
going missing, and of machines crashing and losing stored votes. Only
a few such cases can damage confidence badly: and crashes, at the
least, cause huge delays.

The solutions are not hard to find: a wholesale switch to paper
ballots and optical scanners; more training for election officials;
and open access to machine software. But it is too late for any of
that this time—and that is a scandal.


Copyright (c) 2006 The Economist Newspaper and The Economist Group.
All rights reserved.