I would agree that there's no reason to have e-voting instead of paper ballots. The only possible excuse I might imagine for e-voting is that it lets people vote from home which enfranchises more voters, but that would just lead to a wave of vote selling so...
The Help America Votes Act, which spurred much of this new voting technology, specifically says voting places and ballots should be accessible by voters with disabilities such as the blind. This is very hard to accomplish in a secure, protected way without a machine. Some polling places have paper ballots for most, and machines reserved for the disabled.
As a counter-example of having to trust the receiving end, consider zero knowledge proofs [1]. Even if you don't trust the prover, you become convinced of what they're trying to prove to you (if it's true).
The basic trick to proper cryptographic voting machines is zero-knowledge-proving that they are working correctly. Any tampering, whether it be in the source code or at the hardware level, will either not affect the tally or break the zero knowledge proofs.
Everybody understands paper voting and there are several (more or less) trustworthy protocols for paper-based vote count.
Maybe I'm pessimistic, but I can't picture a near future where the populace understands computers enough to be able to trust e-voting.
All the 'did I vote for the person I tried to vote for?' happens either in the voting booth, or by choosing unused ballots to audit at random.
If vote is secret you just know that someone voted X. Who is that someone? Is he even real?
Since we need secrecy we need to know X, but not who is "someone", so we're blind there.
If you can't prevent fake voters from registering, you've failed before the election even begins.
(Yes, this does allow you to coerce people with known preferences into not voting in order to affect the result. This issue, and many others, are discussed in the video.)
I think a startup to fix voting would actually be a really cool (anyone here doing that already?), although I imagine it would be fraught with danger.
I think we need to step back to first principles here. In an ideal voting system, we need two properties:
1.) The vote must be anonymous (to remove the possibility of persecution) 2.) There has to be some way to detect if a vote has been tampered with.
Here's the thing... we can solve BOTH those problems with common cryptographic algorithms. Problem #1 can be solved with a hashed identity, problem #2 can be solved with a checksum.
Right now we try to solve that with a "paper trail". That's one solution, but it's very problematic. First, you're putting a lot of trust in the people doing the counting. Second, we can't trace those pieces of paper back to people to ask them "is this really who you voted for?"
I'm not claiming these are /easy/ problems to solve. There are a lot of considerations. But I will claim that we already have the required cryptographic toolset to make it possible, and that simple counting machines are entirely inadequate.