[E-voting] The best VVAT is good for
cansbro at eircom.net
Mon Oct 24 17:07:57 IST 2005
I see your point, technically speaking, that a thing cannot be its own
audit trail. But the only way to provide that would be to connect
voters to their vote, and do something like ring up a random sample of
voters and say, "Did you really vote for Xxxxxxxx Xxxxxx #1 and
Yyyyyyyyy Yyy #2 on your ballot?" Which is forbidden by the
Constitution. (It's funny to imagine, though--horrifying as this would
be. Plus the voter could say anything they wanted--what if it was a
voter with a poor memory?!)
In other words, an "audit" in this technical sense would not be
possible. However, with the proposed paperless DRE we'd have /nothing/
to check that a voter has actually seen. The voter can't see the actual
bits and bytes, they only get to see a display.
I wish we could get some kind of hand counting done at the voting
places. I don't yet see any practical way of doing this. As a next
best solution, at least 2 observers should accompany the Gardai with the
ballot box to the count centre.
Brian O'Byrne wrote:
>A VVAT is a little more than that.
>You described something voter verified, but to be an audit trail it
>must be independent, persistent, immutable and possible to reconcile
>against the vote that is counted.
>A DRE machine that displays the vote on-screen and asks the voter to
>confirm the vote before recording it is arguably 'a mechanism by
>which the voter has verified the correct recording of their vote',
>but there is no audit trail.
>To be very pedantic about it the VVPB we currently use is also not an
>audit trail. Information security professionals will talk about an
>audit trail having failure modes different to the thing being audited
>(or better: having no common failure modes), so a thing cannot be its
>own audit trail.
More information about the E-voting