Mimblewimble And Scriptless Scripts
Speakers: Andrew Poelstra
Date: March 4, 2017
Transcript By: Bryan Bishop
Tags: Adaptor signatures, Sidechains
Media: https://www.youtube.com/watch?v=0mVOq1jaR1U&t=39m8s
https://twitter.com/kanzure/status/838480912396533760
Up next we have Andrew Poelstra and he is going to be talking about a really interesting new concept for a cryptocurrency. It’s called mimblewimble, after a spell from Harry Potter, because it prevents characters from speaking. Try speaking? Hello, test. We also have a… hello? hello? Can you guys hear me?
I am going to talk about the concept of scriptless scripts. It’s a way of describing what mimblewimble does, in a way that shows the conceptual details, and maybe give the ability to take the lessons we learned from mimblewimble and apply to other projects.
“Scriptless scripts” are magic digital signatures that can only be created by faithful execution of a smart contract. So what I mean is taking a smart …… if they do this, the result is a digital signature which people can verify. This has some cool privacy consequences.
You might think this is limited in power, and it is, in an academic sense, but there’s a lot we can do with this paradigm. Verifying digital signatures is a very general thing, but it’s powerful. Mimblewimble only supports scriptless scripts. I’ll build up to that as I run through my talk. I’ve talked about mimblewimble before, so maybe I’ll leave it unstated if I don’t.
Bitcoin, ethereum and these different blockchains have a scripting language which is a way of describing under which conditions coins can be spent. These scripting languages allow you to do smart contracts where coins can be spent under conditions like timelock, multiparty, delay, random numbers, or some other …. everybody downloads, everyone verifies…. ings… so they can’t really be compressed or aggregated in any coherent or consistent way. They actually end up being hashed, so they can’t really be compressed. It just doesn’t work, without breaking the hashes, it just doesn’t work. Secondly, the details of these scripts are public, they are visible to everyone forever and they have to be stored. This means different outputs in bitcoin, or different accounts in ethereum, are not very fungible or private because you can tell what the rules are on each individual coin and you can distinguish ebtween coins and track all this cool stuff, because of individual scripts moving coins around.
In scriptless scripts, the only thing that matters is whether the signature is valid. These are basic cryptographic structures. They just look like random data. It erases so much of the public data. It’s also a consistent data structure, so you can have a lot of compression.
Let me do a bit of a digression to explain Schnorr signatures. You have an ephemeral key pair, like k and kg here, you compute this value s which is a simple linear transaction… values.. and the position on the line you choose is a hash of all the data you want to commit to, which is what a digital signature is. The verification is simple, there’s a discrete log problem which is that i can put G’s in there and you can’t figure it out. I can publish all these values and… because it’s multiplication.. equal signs and minus siigns don’t get broken. The verification equation… makes sure the signature was… pretty much anything out there. Just the equation from my early slide, but I’ve added some… from one of them. You can verify this equation the same as verifying an ordinary Schnorr signature, you multiply everything by G… this verification is a weird special property of this value, what it’s verifying is that if this verification equation checks out, then given D, a valid Schnorr signature too, you can figure out the other one, just given the first signature. D is a translating key between two separate independent Schnorr signatures. So to make two transactions atomic, one is happening from one party, the ohter transaction is happening with the other, it’s 2-of-2, each party has a half that they are contributing. During the setup phase, someone gives this other party the value p, and then once they give that value away, as soon as they sign one transaction, that signature can be used to tweak and make that other transaction. You get this atomicity property from hash preimages here. Both lightning and general atomic swaps, there’s only really one party, so they can.. sure that they… And what’s exciting about this value D is that it’s the difference between two Schnorr signatures. So someone is sharing the value p before the signatures are made; once the transactions are out there and public, .. I can take any two Schnorr signatures in the world and get their difference, and I can go make up a transcript. So this value D doesn’t provide any linkages. These remain independent separate signatures. The only magic is that D is public before the signatures are public… this is critical because in mimblewimble we don’t have support for scripts, and it was thought for a long time that we wouldn’t be able to do atomic swaps or lightning network. This was a big deal because mimblewimble scaled pretty well, but if we can’t do opt-in payment channels then this blockchain scalability story kind of get undermines. People would say well bitcoin in principle we can setup indefinitely many transactions in a payment channel, and it’s great that your mimblewimble transactions are small but it can’t handle payment channels. However, using scriptless scripts, we can make payment channels. We can do this on mimblewimble and any blockchain that uses Schnorr signatures. If bitcoin had Schnorr signatures, and I think we probably will in a few years, you could do a closing scheme where currently in lightning everyone has a hash preimage and every channel in the path has to have a preimage of some hash, you could use each linkage in the key to the path to have a different D value, you still have the atomicity and channels and they can only be created/destroyed in one go. You no longer have hash preimages in places that people have to store and validate forever. We get a win for fungibility, privacy, etc. An overarching theme has been what makes a blockchain– well it’s mostly signatures, they all look random, there’s nothing for people to really track here and attaching taint or censoring or whatever. Everything sort of looks the same, which was the goal of scriptless scripts and that we achieved here. I came up with this to get mimblewimble and lightning. As a bonus, it turns out to be very generic.
So let’s get on to mimblewimble here, which I claim is itself an ultimate scriptless script. The way that mimblewimble transactions work.. for those of you who haven’t seen this before, mimblewimble is a blockchain where every transaction has inputs and outputs. Every input and output has a key. Confidential transactions everywhere. In a valid balance transaction you can take the input commitments and output commitemnts, all the pederson values balance out. So this is like a multisignature key. If you take this difference, outputs minus inputs, you get a multisig key of the outputs and a multisig key of the inputs. We call this the “kernel” in mimblewimble. To make the transaction valid, we require a signature with this key. The signature authorizes the transaction and proves that the owner wanted the transaction to happen, and secondly the fact that the signature is possible, which means the transaction balanced, so it’s using this proof of non-inflation really. So what this is is that the validity of a transaction, meaning all relevant parties authorized it, and no coins were created from nowhere or were destroyed. The validity comes down to a single key and a single signature and ultimately everything mimblewimble except for these keys and signatures is just for witnessing what’s going on. The core of mimblewimble is tha tevery transaction is compressed into a single key and a single signature. The signature is simple enough, there’s no real requirement, there’s one thing that we need to add, you need to sign a locktime, to make it possible to back out if people ban, but I’m not mentioning that here. The very simple signature can be used in any of these other things, we could make the preimage the decryption key of something, we can use an atomic swap mechanism, I can do whatever I want, so even though these are sngle signatures, I can enforce arbitrary multiparty smart contracts into these single signatures. That’s the real magic of mimblewimble.
I will end with a few open questions. One is a generic scriptless script; I gave some examples of cool things to do. I wonder if there’s a generic way to do this, like is there a way to do a NAND gate in scriptless scripts to execute arbitary script? The second one is locktime. I can’t do locktimes in scriptless scripts, without SNARKs, which isn’t efficient. Locktime needs to be publicly validated, which as we talked about, scriptless script works with…. so it would be cool that instead of a locktime in the blockchain forever, if we could remove that, as the last wart of mimblewimble. Everything needs to be just keys and signatures. Those are all the open problems, an exhaustive lst of open problems.
And that is my talk, thank you.