As someone who's been involved in the electronic voting controversy, I'd
like to add a few points:
(1) I agree that source code should be inspected by someone truly
independent and competent, and that the standards for approving voting
machines should be stronger. However, that's NOT the same as open source.
And I'd strongly discourage folks from calling for open source, as it plays
directly into the hands of folks like Diebold, who claim that the people
(like me) who want Voter Verified Paper Audit Trails (VVPATs) are really
trying to kill free enterprise. [Yes, I know all the examples of businesses
based on open source, but that's not what this is about.] As an example,
Harris Miller, the president of ITAA (www.itaa.org), a politically
influential consortium of technology vendors, is on record as having equated
the VVPAT groups with the open source community. So rather than putting
your energy into trying to get Diebold et al to move to open source, it
would be far more productive to put your energy into VVPATs. Towards that
end, I'll encourage everyone participating in this discussion to look at
www.verifiedvoting.org. VVPATs can give us the assurance we need of
accurate elections, without delving into the political morass of open source
and related topics.
(2) WRT the web page showing a "Sun server when discussing Windows", I hope
people realize that web pages for companies are made up by marketing people
who don't understand the difference. Don't hold that against them... There
are plenty of real reasons to oppose Diebold.
(3) WRT requiring that the technology protect itself in case the users
don't, that's simply unrealistic. In *any* real computer system, there are
expectations about the environment (e.g., the administrators aren't hostile
to the functioning of the system). It's important to state what those
expectations are, but there will ALWAYS be some that rely on non-technical
means. The important part about election systems is that they be explicitly
stated, and they be enforceable using non-technical means (e.g., by having
locks on doors). The problem today is that some of the assumptions (e.g.,
the vendor provided software doesn't have any bugs) are clearly unrealistic.
(4) WRT getting one set of software approved, and then installing another...
that's an old problem in any environment. The way it's supposed to work in
election systems is that a particular version is approved, and it's illegal
for the vendor to install something different. If there are teeth in the
law, and the vendor can be fined for installing illegal software, then it's
a reasonable non-technical measure. Of course, one could also use things
like cryptographic checksums to verify that what's installed is what was
approved. That still requires non-technical elements, such as that the
people who ran the checksums weren't deliberately trying to cover up a
change, the checksums were protected from tampering, the software that
calculated the checksums wasn't subverted, etc. [For those of us old enough
to remember, vendors were required to address this as part of "Orange Book"
evaluations, and are now required to address it as part of Common Criteria
evaluations.]
Bottom line, election systems are no different than any other systems in
that the security of the whole system is based on risk management. While we
should have higher expectations of election software than office automation
software, let's recognize what it is. IMHO, VVPATs are the only real way to
go.
--Jeremy