I am just theorizing here: Someone now takes the box of paper votes and runs it through the scanner machine. And passes this number along to someone. What is stopping them from tampering at this step? I think this is precisely what my co-worker was describing. There is an inherent trust that your paper ballot is scanned and recorded in a fashion that matches your vote.
An electronic voting machine could potentially communicate votes in real time over a secure connection. Or in the case of Brazil's machines, I believe stores it locally, encrypted, with a verifiable cryptographic signature of some sort.
I'm sure we all know the multitude of other attack vectors this introduces. I guess I am just not convinced that paper makes things more secure.
We designed a vote printer that would allow the voter to see a paper copy before storing it, but it was never used.
Given the number of comments you've made on this thread, it seems this is an area of interest to you. I encourage you to look though the previous HN discussions on this topic. Here's a list of some of those from the past month or so:
Doesn't scale is a concept that only developers and entrepreneurs understand. Sadly, that makes it an invalid argument for the other 98% of the population.
Where I vote, my paper ballot in no way identifies me. I identify myself upon entering the polling station, they find my name on the list of registered voters and mark it. When I'm turning in my completed ballot, I again identify myself and my name is marked on a separate list. So there's a record that I voted but not for whom I voted. How would an electronic voting machine improve upon this?
BTW, where I vote, the paper ballots are the bubble scan kind and the voter feeds it to the machine themselves. This provides very fast tabulations with a paper record for security and recounts.