Placeholder Image

Subtitles section Play video

  • Asked me to talk a little bit about the

  • new innovations in voting systems. And so I'd like to tell you about end-to-end

  • verifiable voting systems,

  • which is the buzzword that academics are using to describe this.

  • Now,Which means really that you can verify the whole flow of

  • information, from your head, as to who you think you want to vote for,

  • all the way through the casting of the vote and the tabulation

  • and then the final tally,

  • so that you can verify that every step of the way is going the way it should

  • and that your vote counts the way you intended,

  • that the final result is the accurate representation of everybody's votes.

  • And right now, with many voting systems,

  • you can have a certain amount of verifiability. The lovely thing about paper ballots

  • is that you can see that, you know, what you wrote down is what you intended. And so that's a

  • very important property, very nice.

  • But once you've cast the paper ballot, you're trusting the chain of custody of those paper ballots

  • to make sure that the count is done right

  • and that the ballots being counted are the right ones.

  • Brady: "Talk to me about this selfie thing. Some people like to take a photo of their ballot, don't they?"

  • Yeah, it's a lot of fun, and I think it may increase voter participation

  • to have a little more fun like that in the ballot booth and so on, too.

  • The thing that is worrisome about it, though, from

  • a security point of view, is that, you know, with a selfie of you casting your ballot,

  • you can sell your ballot. And that's one of the things that we've learned the hard way over many many decades,

  • is that you don't want to enable people to sell their ballots.

  • So you shouldn't leave the voting booth with any kind of proof as to how you voted.

  • You shouldn't have a receipt saying you voted this way, you shouldn't have a copy of your ballot,

  • you shouldn't have anything that proves how you voted.

  • Because that can be worth, you know, 20 bucks to the, you know, to you,

  • from the guy down the bar who wants to, you know, buy your vote. Or maybe coerce you.

  • So, yeah, how can you possibly know that your vote was counted the way you wanted?

  • This is a hard question, and it makes a fascinating research area. And it turns out

  • you can thread this needle with a bit of care and a bit of cryptography, a bit of mathematics.

  • So how do you do this?

  • So the idea would be that you get a receipt when you vote.

  • And the receipt is not the plain text as to how you voted

  • but an encryption of how you voted. So it's something that you can't sell down at the bar,

  • because it's not something that's interpretable by somebody else.

  • Brady: "So what is it, like a number, or a barcode, or...?"

  • It's a number or a barcode or something like that. It represents how you voted.

  • So you get a receipt,

  • and, the new thing here in these end-to-end verifiable systems

  • is the copies of all the receipts get posted on the web.

  • So say, Brady, and here Brady's receipt, and Ron, and here's Ron's receipt, and so on and so forth. Everybody's name and their

  • receipts get posted.

  • Nobody can tell how anybody else voted. You can't sell your vote based on that posting. But it's all on the web,

  • and it's all there for anybody look at. And you can look at it, and you can see that, yep,

  • they've got my receipt right, and that receipt is an encryption of my vote, right?

  • So that's the key thing. You know your vote is in the pile,

  • and you know your vote is in the pile the way you think it should be,

  • because you have reason to believe, as I'll explain,

  • that your encryption of your vote is actually correct.

  • Brady: "Everyone's encryption looks different, though."

  • Everybody's encryption looks different.

  • So it's not a simple encryption in a deterministic way. It's a good point.

  • Yeah, so if everybody's voting for Alice, all the Alices don't look the same, and all the Bobs don't look the same.

  • So there's a couple of subtleties here. One is, well, how do you verify the tally then? So somebody is announcing that

  • Alice won. And so, all these encrypted votes are

  • primarily for Alice, and there's some for Bob, but there's more for Alice.

  • That's the assertion that the election officials have made. How do they know? How do they

  • convince anybody that that's right? And they'd like to convince everybody that that's right without showing the decryptions of everybody's vote.

  • So they've got this pile of encrypted votes and an announced result,

  • and you'd like to know,... if you're an election official, you'd like to convince everybody that's the right result.

  • So there's a number of ways of doing this, and there's quite a large literature on this. Here's a way

  • that's pretty simple to describe. So you could use something called homomorphic encryption.

  • What does that mean? So it means that you can multiply

  • ciphertext together and end up with them representing the sum of their plaintext.

  • So, encryption of a one times an encryption of a two will give you an encryption of a three, right?

  • So you end up adding the plaintext when you multiply the ciphertext. So this is cool.

  • You can do lots of wonderful things with this. And in particular, you can do the tally for the voting system.

  • For example, if your vote is a vote for Alice, you might have an encryption of a one.

  • And if your vote is a vote for Bob, you might have an encryption for a zero. And suppose Alice and Bob are the only two candidates.

  • So then, each ballot is either a one or a zero, and basically what you want to do is add them up.

  • But they're ciphertexts, so you don't see the ones and the zeros.

  • But you can use this homomorphic property, you can multiply all these ciphertexts together

  • end up with a number which represents the sum of all of the

  • plaintext ballots. And that sum is just the number of votes cast for Alice then, right?

  • So the tally for Alice is the sum of all these things. And Bob is just the number of votes minus the tally for Alice,

  • of course. So if Alice has more than half, she wins.

  • So there, you can multiply these ciphertexts together and get a ciphertext that's the ciphertext for the sum,

  • and that ciphertext can be decrypted publicly

  • in a way that everybody can see is proper. So you can see yes that is... and everybody could also do this multiplication,

  • So that is, they can check that that's the ciphertext to decrypt,

  • and the election officials could decrypt that in a nice way

  • and show that that's the proper decryption, so that we have the right tally for Alice.

  • So that's one part of what you need to do then, is everybody can verify the tally from these encrypted things.

  • And that's cool, right? So that's one part of it. The other part you want to

  • deal with is the fact that you need to know that your ciphertext is really representing your vote, right? So you go into the

  • voting booth, you're getting this receipt,

  • and you want to know that that receipt really represents your vote. So you vote for Alice, and you get this receipt which is a

  • ciphertext,

  • you can't decrypt that ciphertext. In fact, you shouldn't be able to decrypt that

  • ciphertext, because then you could sell your vote, just like you could with a selfie, right? So you want to know that that

  • ciphertext is really for, say, Alice, if you're voting for Alice.

  • So, that's a little more subtle.

  • Maybe this matchbox analogy is not a bad way to explain it. How do you

  • know that you've got a good match? You're going on a camping trip, you can, you're allowed to take only one match

  • How do you know you've got a good match?

  • Well, you could take a box out of the drawer, you can say well,

  • maybe this is a box of good matches.

  • And you try randomly sampling those matches, and, light, yeah, that one lights, that one lights.

  • But you want to take with you a match you haven't tried yet. And so you,

  • maybe the last match, you say, well, I haven't tried that match, but all the others in the box were good, and therefore

  • I'm going to trust that this one is good. You put that your pocket and take it camping.

  • So you can do something like that with the encryption here. You can say, I walk into the voting booth,

  • I get an encryption of my first

  • representation of the choice. I say Alice and I get encryption, then I want to see,

  • is that really a good encryption? It's like seeing if it's a good match.

  • So I'll strike it, or I'll ask that to be decrypted,

  • which the machine can do for me, since it

  • encrypted it in the first place. So, so once I test it, I, once I see that, how it decrypts properly, unfortunately,

  • I can't use that to vote with, because now I know how to, how it decrypts as well,

  • and I can't post that, because I could sell my vote if I were to post that one.

  • So that's a spoiled ballot, and, and we can't use that.

  • Brady: "That gets taken off the tally."

  • It's taken, it's not even posted,

  • it's not in the tally, it's not used, it's not anywheres. It's like I haven't voted at all,

  • I'm just testing the voting machine. And I can do that as many times as I like, you know, two or three times,

  • check out different decryptions of different things, and I can see the machine is reliably encrypting

  • whatever I say I want to vote for. I can say I'm voting for Alice and test that, I can say I'm voting for Bob and I test that,

  • and say I'm voting for Alice again. Maybe I don't test that one, and I say, okay we'll go with that,

  • we'll post that on the web. And then you've got confidence, just like with the matches, that that encryption was done properly.

  • It doesn't need to be done by a lot of people, right? It just needs to be done enough

  • so that you know machines aren't trying to cheat in any large-scale way.

  • So those are the two parts of it. Getting encryptions that you can trust, even if you're not a computer,

  • and getting a tally that you can verify, even though the inputs of the tally are encryptions.

  • And so this is a wonderful idea. It's evolved over the years with many different people

  • putting parts of it together, but it means you can have a voting system, then,

  • that's verifiable, all the way from who you want to vote for, that end of the process,

  • through the casting of the ballot, through the posting of the ballots, through the tally and the announcement of the result. You can verify that the

  • outcome is right. So, in the future, this might be the kind of voting system

  • everybody will vote on. But it's still in research stages a bit. It's under development.

  • There are some jurisdictions that are starting to use it in practice, and we'll see how it goes. I'm optimistic.

  • ...Alice,

  • so put another Alice in, pick one out, and put it back, together with a copy of it. So now I've got

  • six ballots.

  • That's cool. So now I've drawn the bits. Now it's four to two...

Asked me to talk a little bit about the

Subtitles and vocabulary

Click the word to look it up Click the word to find further inforamtion about it