It would make more sense to check whether your vote was counted post-facto but it's hard to especially if you still want to make it difficult to figure out who voted for whom. There are a few mathematical possibilities but they don't seem particularly practical to me: https://www.youtube.com/watch?v=BYRTvoZ3Rho
So for the time being paper ballots are probably the way to go.
You vote, confirm the receipt says the right thing, and stuff it in a box. Verified hardware is a solved problem. Given that it'd have no idea of the candidates, there's no way to program it ahead of time to favor a candidate without being caught.
We already have election officials watching each other. This needn't be more complicated than a late 70s calculator. It can even be so simple it doesn't know the candidates names or party affiliations.
Well, in theory... In reality, it will be a black box, connected to any open wireless, and made by the company who donated the most. But, in theory, it'd not be too hard to make a fairly dumb device that provides verification. It only has to count and print.
That's fair. I guess you could also force a recount for a random portion of the polling stations to detect anything fishy. I'm not sure if it'd be really worth it though, would it be such a massive improvement over simply counting paper ballots directly?
>Checksums and verified hardware.
Nah, that won't work. Who checks that? Who makes sure nobody tampers with the hardware after it's been verified? How can you even verify a piece of hardware, it could tell you it's running code X when really it runs code Y, no amount of checksumming will help. You could easily backdoor a processor to run a special ROM instead of the official one. Good luck detecting that.
Are we supposed to send electronic engineers at every poling station to check for any tampering with the hardware? How about people making copies of the chips that look exactly the same from the outside but are backdoored? You can't just trust whatever is printed on the package.
So what do you do? Take a random sample of machines during every election and decap all the chips to look for something abnormal?
IMO it's just too much hassle for very little benefit.
In the end a trusted minor miscount would be less bad than a count that is widely untrusted despite being correct: one would mean living through a "what if some people would have cast a different vote" alternative reality, the other would be a full blown legitimacy crisis, with unconstrained cheating in the next election as the least bad of all possible outcome.
Just stick to paper ballots.
We can put people in space. Verified hardware and software is a solved problem. Well, assuming you don't make it absurdly complicated. All it needs to do is eliminate manual counting - while providing a paper printout should someone need a recount.
You don't even need special tokens to make sure the person in the booth is only voting once. You can be as simple as making it flash an LED to be monitored by the election officials. They have volunteers from the various parties already there watching. Watching a light may even be easier.
It doesn't need to be complicated. It need only be a glorified abacus.
Separate from that, what does using millions of one-use physically collected chips buy you that just writing to a paper ballot doesn't?
You buy millions of chips so that we never have to go through this again. This way, the validated code and hardware don't have to be worried about, for a long time. The chips go in the machine, they run it. You'll have spares for a very long time.
What does this do? When the polls close, they don't have to count them. They need only count if there's a demand to do so, like a recount today. That's it. Problem solved.