The reference point is [Trusting Trust]. It was his Turing Award lecture in 1982. Honestly, it and similar materials should be required reading somewhere in university CS and Software Engineering curricula.
> I nervously laugh whenever somebody talks about "open source electronic voting"
Uh, why? I don't think anyone has seriously asserted that open source voting systems would be foolproof. On the other hand, we have actual cases where open source voting systems would have at least made serious flaws public before they were used for voting.
Besides, when you're talking "Trusting Trust", you have to worry just as much about the hardware you're voting on. Who designed and built that?
Also, don't replace a secure, well-tested and practically proven system that works (paper and a ballot box), with something that fosters large scale malice (electronic voting) in the name of modernizing or, even worse, preventing small scale malice.
Voting is a process with unique constraints not seen elsewhere. Not with banking, not with aviation, you name it. Mainly because each voter must be able to cast his/her vote and be able to verify if the vote is counted correct and exactly once without letting anyone else know what he/she voted for.
With paper and ballot box, you need to trust the people doing the counting.
And since you usually need a lot of them when it's done manually, it's very probable that there's always some amount of fraud.
Now, it's probably easier to fraud if you can hack the voting machines. But let's not say that the current system really prevents fraud because it doesn't.
Hence my argument about large scale fraud (only possible with electronics) vs. small scale fraud. Fraud with paper and ballot boxes is extremely hard to organize at large, because it would involve a couple of orders of magnitude more people to conspire.
The electronic voting has its advantages and given the context of the discussion about trust and security, your laugh is related to the means of ensuring such security, not to the idea of electronic voting. The means should and can be treated separately.
If you must have assistance for people who can't mark a ballot paper due to disability, have the machine mark their ballot paper and count it as per normal.
Everyone knows what cheating looks like in ballot counting, the evidence of ballot stuffing is harder to hide, easier to spot and generally understood.
Even if electronic voting could be made absolutely secure, safe and fair it would not be understood by most voters, that means it really shouldn't be used. But as pointed out, it can't be made secure so just don't use it. Ban it. Destroy it. Elections must be fairly counted, seen to be fairly counted and understood how they are counted fairly.
[Trusting Trust]: http://cm.bell-labs.com/who/ken/trust.html