I’m at the Electronic Voting Technology / Workshop on Trustworthy Elections get-together (in Montreal). A few thoughts about the day and talks.
Keynote
Larry Norden from the Brennan Center is proposing a few ideas about what the voting community should focus on. Voting machine security, including forensics. Thinks statistical analyses to detect fraud is going to become increasingly important (e.g. Iran). Doubts that election officials in the US will care until there’s a big meltdown. The battle for paper-based voting records has been won in the US (Holt Bill). Some resistance to paper because of cost. Need a convincing cost analysis of conversion to optical scan machines. The policy debate about post-election audit has also been won (again Holt Bill). Analysis on cost of post-election audits would be very helpful. Risk-limiting audits is a huge topic (Hall/Stark paper at this conference.) Long way off from getting rid of DREs entirely. Would like to see an expanded role for EAC, e.g. notification hub for problem reporting.
Larry moves on to thinking about evolution of how people voted: vote by mail. All Western States have no-excuse absentee voting. We’re getting to 100%. In-person early voting has more moment (me: well, that’s a good thing, in-person is much better.) Larry is initially concerned about issues unrelated to coercion, e.g. placement of voting sites, securing of voting machines. But then he brings up “privacy issues.” Postal workers handling ballots. Usability issues of vote-by-mail, double-envelope, instructions, deadline, etc. WA state: 1.9% of vote-by-mail ballots rejected. NC: 9.1% (!!) It’s too late to say “let’s get rid of vote-by-mail.” There’s been a big push for Internet voting, framed around military needs, of course as a “foot in the door.” 14 States have proposals for Internet Voting pilots. The issue of military voting is a legitimate issue. We have to do more than say “it’s a bad idea,” we have to provide an alternative. Registration is a big issue, given military relocation. Registration is a big issue in general. US has a much larger registration problem than most other modern countries. Brennan Center actively working on “Voter Registration Modernization.” Registration should be automatic, opt-out. Leverage existing databases, similar as in Canada. “Registration follows you when you move.”
Norden encourages talking to election officials, institute pilots, etc. [Couldn’t quite tell if he was knocking crypto-based voting or telling people to take it more seriously even if not applicable to the very next election.]
Josh Benaloh brings up the coercion risk of vote-by-mail, which Norden didn’t mention. “The horse is out of the barn door,” says Norden.
Refereed Talks
Michael Byrne talks about how voters notice and react to screen anomalies. Gets lots of groans when he says that there’s no evidence of actual security breaches in voting systems, but quickly corrects himself and says he just wants equal time for security and usability. Half of voters notice review screen anomalies. Time spent on review screen is significant predictor. Voters are aware of how carefully they checked the review screen. Residual vote rates may not be useful [we knew this, I think], because wrong-choice errors are the least detectable and they’re not picked up by residual vote rate. Suspects VVPATs are useless.
Sharon Laskowsi talks about the importance of poll worker documentation / manuals on election day.
Sean Peisert discusses forensics applied to voting machines. Points out the conflict between auditing and enforced ballot secrecy. VVPATs are not forensic audit trails.
Inés Levin discusses auditing of the Venezuelan election purely by analyzing public records.
Benlog 