E-Voting Report Finds Problems with Modern Elections

JonRob writes "The Open Rights Group has released a report on challenges faced by voting technology. Using the May 2007 Scottish/English elections as a testbed, researchers have collated hundreds of observations into a verdict on voting in the digital age. 'The report provides a comprehensive look at elections that used e-counting or e-voting technologies. As a result of the report's findings ORG cannot express confidence in the results for the areas we observed. This is not a declaration we take lightly but, despite having had accredited observers on location, having interviewed local authorities and having filed Freedom of Information requests, ORG is still not able to verify if votes were counted accurately and as voters intended.' The report is available online in pdf format for download."

Comments

Voter-verifiable counting

[peacefinder]

For your consideration, may I present my[1] idea for a voter-verifiable counting system:

---

In addition to any other vote-counting or verification system, a county
elections office could take a full optical scan of the ballot papers.
The data from these scans would be made available to all who request it;
anyone could acquire the data and perform their own re-count with any
method of their own devising.

This would provide complete transparency for the automated portion of
the counting process.

The problem with optical-mark scanners, of course, is that the
scanner's internal software and firmware is vulnerable to tampering.
Such a tampered machine cannot change the ballots it reads, but it can
misinterpret them.

By providing a raw image scan to the public, we'd be enabling many
eyes to provide their own interpretation of the ballots. Any
optical-scan vulnerability would become moot. We would go beyond a
voter-verified ballot, and get to a voter-verified count.

This is technically achievable with commercial off-the-shelf hardware
for well under $100,000 per county in capital expenditures.
Specifically:

* Industrial scanners of sufficient reliability are available. At my
workplace we have a "light" duty commercial scanner with a duty cycle
of 8,500 scans per day; this machine cost around $7,000. If county
clerks were to have about 5 days to produce the scans, two of these
scanners could completely scan the ballots for all but the largest
counties. And, of course, heavier duty scanners are available.

* Since industrial scanners are not optimized for ballot reading or
even optical-mark recognition, it would be much more difficult for any
malicious entity to successfully tamper with their software to produce
inaccurate ballot image scans. It's much more difficult for software
to produce an incorrect image than an incorrect interpretation of an
image. What's more, these scanners are available from several
manufacturers; if one distrusts any or all scanner vendors, one could
simply scan the original ballots with a variety of different
manufacturers' scanners and compare the results.

* For the standard optical-scan ballot, a fax-quality scan would be
sufficient for a voter-verified count. Better scans are possible for
higher time, money, and data storage budgets, but I don't think they
would be necessary as a practical matter.

* The data storage requirement for an approximately fax-quality scan
of every Oregon ballot - approximately 2 million ballots with 100%
turnout - would be under 500 gigabytes uncompressed per statewide
election. (And ballot scans should be highly compressible even with
lossless and error-correcting algorithms.) Portable hard drives that
large are available for around $300. Most individual county ballot
scan datasets would even fit on larger iPods.

---

This brings up a couple other problems, of course. Foremost, the ballots have to be on ADF-feedable paper, and probably had best be marked ballots rather than punched-paper. Also, the question of what to do with a voter-made distinctive or identifying mark on the ballot needs to be addressed. (Distinctive marks could lead to buyer-verified vote buying.)

But still, it's a huge step beyond just trusting the county's optical-scanning ballot interpreter.

[1: Actually this is my brother's idea, which I have modified slightly.]