I'm writing this post as I travel by train to Vienna for a few days of CSF (http://csf2014.di.univr.it/), where I'm looking forward to an invited talk on logic and voting by Veronique Cortier and one one EasyCrypt by Gilles Barthe. The hope is that one day, verifying a voting protocol will be a purely mechanical task and the debate about whether a protocol is secure (and what this means in the first place) will be settled.
But we have a bigger problem to solve. For a few years now, we've proudly been citing Norway as an example of where e-voting is used on a national level. Unfortunately, Norway has now abandoned e-voting due to concerns about the security of the system. Not that anyone has broken it - but a good scheme must do two things, it must offer security and it must offer a perception of security. The two are not the same. Sure, an obviously broken scheme won't be perceived as secure either when people's votes get leaked. But building a scheme that no-one has broken to date or even found a hint of how one could break it, as we've seen in Norway, isn't enough to make people comfortable. Looking at the schemes I know of, it strikes me how much we've concentrated on the former at the expense of the latter - how many cryptographers working on voting have even tried to validate their ideas with tests "in the field"?
For example, one of the mantras of cryptographers working on voting is verify, verify, verify: voters can verify their ballots as a protection against fraud, they can verify each others' ballots which goes some way towards preventing ballot stuffing and independent parties can verify elections, preventing the authorities from cheating on the result. Except that when Olembo et al. ( http://link.springer.com/chapter/10.1007%2F978-3-642-39185-9_9 ) actually asked German voters what they thought about a verifiable system, it seems no-one could be bothered to verify. Why? Because voters' mental model is that if a system is good enough for the authorities to approve, then presumably it's good enough that they don't have to check it themselves! And if voters don't verify, we lose our supposed safeguard against someone manipulating the election. Perhaps this suggests that we should look at alternatives to voter-verifiable schemes, which would represent a major shift in the focus of cryptographic voting research.
Papers like this tend to have a hard time in academia - supposedly not "hard" enough science? But in my opinion, this is exactly what we should, nay MUST be doing - cryptographically, e-voting is a solved problem. Homomorphic ElGamal has been around a while and we've got decent mix-nets too. What we don't seem to have is a fully worked out scheme that could be used in a nation-scale election, or (so it seems) the motivation to build one. With Norway's withdrawal, we now have one less example to cite of why our research is supposedly practical. My personal feeling, as the raeder may have guessed by now, is that we should be spending more of our time on focusing on what a secure voting scheme should look like in the eyes of the general public, and then go away and build something that satisfies those criteria. The impact factor we'd get out of a fully working voting scheme would tower over anything we're doing today.
If cryptographers need any more motivation than that, I'd suggest that if we don't come up with the first real* crypto-voting protocol, someone else will - and they'll have a better sense of salesmanship and business than us, but less understanding of cryptography. And what that will produce, I can only speculate - Diebold Voting Machines certainly sold well.
*Helios is real, but specifically designed NOT for political elections.
I guess that's why I'm on the train to Vienna now: although a cryptographer by trade, the more I can work with people in the more general field that comes under the umbrella term of security, the more I'm likely to discover what the real open problems are. My work with TU Darmstadt so far has been very enlightening in this area. So while I'm very much looking forward to the upcoming talks on logic and verification, I don't think that the next big thing in voting will come from there.
Post a Comment