I take "secure voting" to mean a process where every participant can later individually audit that their vote was counted without betraying the confidentiality of any other voter.
For that, what would you think about a system where the election board published all the votes after the fact without any names, but placing each vote alongside a secret, a key, shared with that voter?
So, as a voter, I could perform a few operations on this published list of results. (1) I could search for my row indicated by my shared secret key to confirm my vote was recorded correctly. (2) I could add all the votes to confirm election officials added all recorded votes. (3) I could confirm that the number of rows is equal to the turnout reported on the day, possibly by district, so no ballot boxes could suddenly be lost or added after the fact.
> If the software is malicious, it can receive one input and produce a different result.
Sure, but all of cryptography is about solving for bad middlemen. The trick to secure multiparty computation is forcing the system to show enough of its work to verify, like above. Just publish the votes with the individual voters strongly masked.
Part of me worries that any voting system without public key cryptography is insecure, but our standards are so low we're just settling for paper, assuming that's the best we can do.
But I'm not sure what attacks paper advocates are trying to prevent, so I'm probably missing something. If I think my local election officials threw out my vote, how do I use my receipt to detect that from home?
But I agree with you if your main points are: 1) Current electronic implementations are terrible. 2) Paper never hurts.
N.b. - Vote selling could be an issue with either paper receipts or my protocol above. The best solution for either is probably outside the system (criminal penalties and significant whistle-blower rewards). You might be able to add deniability to the toy protocol above, but things would get complicated.
A good summary of the problems and how to solve them is here:
If we can't do that, then no system can guarantee us anything. If we can do that, then the system doesn't matter.
I'm familiar with VV. I completely agree with their position that post-election audits are critical to ensuring trust in the system. (That's a weaker version of my own requirement above.)
However, their latest policy recommendation was that we must keep hand-marked ballots, which is just bizarre to me. That's the sort of policy you endorse if you want elections to be decided by teams of lawyers debating stray marks and creative misspellings to deduce voter intent. It's a good way to let money break ties in elections, and makes me question VV's impartiality or judgment on these issues.
They have repeatedly linked to outside teams of researchers covering end-to-end verifiable systems though. A thorough read of the research notes that while there are a number of challenges to overcome, end-to-end verifiable could provide a path to secure elections.
The downside of that is that such systems generally enable bought votes and extorted votes.
> Vote selling could be an issue with either paper receipts or my protocol above. The best solution for either is probably outside the system (criminal penalties and significant whistle-blower rewards). You might be able to add deniability to the toy protocol above, but things would get complicated.
Extortion and bribery work against individuals some of the time, but using it against large groups of people hoping not even one of them will alert reporters or police?
Not to mention, this is a flaw in the current voting system, because people can easily sneak phones into booths to take video of their votes.
Being able to confirm that your specific vote was counted should be considered a basic human right. If you look at it that way, it feels strange to deny people a basic right to protect against some harm we don't currently address now, but could easily address in other ways.
No, not if that means that many people cannot safely vote. At one time it was considered a right in many jurisdictions that everyone could know what everyone else's vote was, and votes were public. This enabled the local landlords to ensure that bad things happened to people who didn't vote the "right" way. Many vote-confirmation systems allow others to verify the vote as well, and the historical record makes it clear that this can be very dangerous.
Currently we can't even verify that votes are counted correctly at all. We currently allow people to press buttons and then have an unverifiable machine report what its owners want the vote to be, without any reasonable way to verify it.
It's great to want much more, but let's start with the basics. Having a way to verify that the counted vote is the actual vote is that basic.
Election security is certainly more than software or hardware. But I'm sick and tired of jurisdictions that permit paperless votes, because by definition they are insecure.
> You can mitigate software flaws with chains of custody, tamper proof seals, bipartisan poll watchers, bipartisan election judges, and enough transparency.
Not true; none of these mitigations work. If the software is malicious, it can receive one input and produce a different result. Reviews of source code and even machine code are not enough, because you don't get adequate confidence that what was reviewed is what was run. "Tamper proof" seals don't help, because the malicious code could have been what's installed in the first place, and hardware can easily report "what you expect to see" while running something else in practice on election day. In most cases the source code isn't even available for public review & scrutiny, so there's no possibility of enough transparency even if you could somehow be assured of what code was run.
When I was a kid I liked to do magic tricks. I bought cheap plastic devices that looked like they did one thing, but in actuality did something else. Modern DRE systems are just a magic trick device - they'll do something, but not necessarily what you think they do.
> Most election systems in the United States employ all of these controls, which is why you have few stories of legitimate votes being discarded due to software failure or tampering.
You hear few stories because there is no effective logging, and therefore no way to report failure or tampering.
> If your argument is that paper ballots are more auditable or harder to hack, they are not. The same access controls above can be applied to paper, and paper is always susceptible to being destroyed, lost, or conveniently found in a clutch race.
Paper ballots are subvertible, but because they are physical items someone has to physically be there to do the manipulation. There's also a long history with paper ballots, so the ways they can be subverted are known - as are their countermeasures.
> Hackable machines are a problem, but they are not the only problem. ...
It's true that these machines are not the only problem, but voting systems are systems. For voting systems to be legitimate we must address all parts. Direct Recording Equipment (DREs) subvert the entire voting system, and thus should never be allowed to be part of one.