Voting by paper as described is how it's done in most western countries for years and years. In general there are very few accusations of voting fraud.

This is because - among other things - that the system is transparent to everyone involved.

You are trying to fix a problem which don't exist - and you are fixing it the wrong way. E-voting opens up a huge amount of possibilities of cheating and worst of all it is going to be even more difficult to find out when it happens.

Worst of all the entire process will end to be transparent to the layman...

I agree. Voting is not a technical endeavor. It's a human endeavor. Trust.

I suspect it might have the added benefit of putting some of the partisans in the same room and making them work together. It's hard to vilify someone to their face, and/or when you are working with them for the entire day.

As for the convenience factor. If we are truly to govern ourselves, surely that endeavor merits a bit of time and effort. To borrow an aphorism, "Penny wise, pound foolish".

We should be sure to make time for people to engage in the process, and to ensure to their satisfaction that it is fair.

Creating boxes filled with votes (and swapping them out for actual ballot boxes) would be a lot harder than you'd think. As someone who's spent a lot of time scrutineering elections, here's how the day runs down.

1. I show up before the polls open, sign myself in, and have the opportunity to look inside each of the ballot boxes. Once I (and other people from other parties doing this) all agree they're empty, the boxes are taped up and the polls open.

2. Over the course of the day, people come and go and vote. At any time, I can continue to be sitting and walking around the room supervising things.

3. At the end of the day (when the polls close) the deputy returning officer closes the doors, leaving only the election workers and any observers inside the room. At that point we (the DRO and the scrutineers) go from table to table verifying that the tape is intact, cutting it open, dumping the pile of ballots on the table, and verifying that the box is empty.

4. The clerks for each poll count the ballots. It's rare for there to be disagreements on who a vote should count for (all ballots look like http://www.elections.ca/content.aspx?section=vot&dir=yth..., with that circle on the right being about an inch in diameter), thought it happens. When it does, we (the representatives of the parties) discuss who we think the vote was for, but ultimately the poll clerk decides. Generally one of the partisan observers objects to this, which his noted in the log.

5. Once all the ballots are counted, the poll clerks fill out a mathematical worksheet to make sure all the numbers (ballots counted, electors checked off, unused ballots, and ballot stubs) match. About half the time these numbers don't, so we recount everything a few times until they do.

5. Each partisan representative gets a carbon copy of the official return form (specifying the result of the poll), and witnesses the DRO calling the results into the regional office.

6. Everything is packaged up on a per-poll basis (used and unused ballots, paperwork, etc.) into a nested series of plastic envelopes, which everyone signs along the seals. These are kept for a few years for auditing.

One of the main benefits of this system is that it's dead-simple, and the counting can be explained to anyone - which is something that many proponents of cryptographic voting don't sufficiently take into account. As for making ti difficult to trace back where fraud took place, I'd say that's almost a selling point. If you could work backwards to exactly where someone faked something, it seems like it would be only a small step further to be able to work out who people individually for.

What happens if someone covertly breaks the seal on a box with the intention of simply invalidating the ballots inside? In some areas, votes tend to go one way or the other depending on the time of day. You might be able to exploit this fact and attempt to invalidate boxes that favor a particular party.

Yep, this is an endemic problem with end-to-end verifiable elections everywhere: it is easy to invalidate whole segments of votes. For example, in the wombat voting scheme, if one of the people who holds a decryption key refuses to divulge it, perhaps by "losing" their key, they can invalidate the whole election. Typically to minimize the chance of this happening you make sure that everyone with a key has a strong commitment to the election and make not divulging the key a public shaming event. "OMG, the Democratic Party refuses to decrypt the vote in this Republican-dominated area -- they must be corrupt!"

It's an awfully sticky situation, because the A (availability) in the CIA principle is so critical in elections, where it typically is compromised the most in encryption schemes. (i.e. Quantum Key Distribution ensures that Eve can't listen in on communication, but doesn't do much to stop Eve from making communication impossible).

Wombat is using threshold cryptography. Threshold cryptosystem, works like this (in short), in order to decrypt an encrypted message a number of parties exceeding a threshold is required to cooperate in the decryption protocol. Meaning, before the election has started a threshold is set, let's say there are 4 candidates, you can agree that 3 of the candidates is enough to decrypt the votes. Also... Don't forget the paper backup, you can simply count those.

This is what I DON'T like about paper and pen methods of voting: "About half the time these numbers don't, so we recount everything a few times until they do." There are errors. How can you guarantee that when the numbers finally do match, that there still isn't any errors on the count? Maybe +1 was added to a candidate, -1 for another, and now the numbers match.