Pablo Villalba - The potential of DVT in Permissionless Liquid Staking

Transcript

welcome back everybody thank you for joining I am very excited to introduce Pablo one of the co-founders of Diva and he is going to be talking about uh liquid staking permissionless liquid staking and DVT so give it up for Pablo so thank you everybody for coming today I'm going to be talking a little bit about how liquid staking and DVT intersect in a in a way I think is very very interesting it creates new possibilities so what Diva staking is doing is is building an Unstoppable liquid staking protocol where you have the all familiar liquid staking on one side you can use it just to get your LST wrap it and not rasing not rasing but

the main difference here is that in Diva all the operations are powered by permissionless operators that anybody can run so anybody can run our node from one e and that's using DVT to kind of uh support each other and cut cut the possible risk of validation um and for that it uses distributed cryptography and it's secured by collateral so the main things he can do with diva is you can stake your eth um and hold it to get your network rewards say 4% you can use that D on defi in a liquid way or you can lock it to run a node and this kind of more Rovers or more like kind of um running a node kind of mode so where we

are today is device prelaunch um we recently launched an early sticker program which attracted for $2 million in tbll there is a test net running I'm going to show a little bit how that works and there's also an active da which has been like helping curate the direction of the protocol so um diva is designed to be a little bit different than everything else so it's designed to avoid every point of failure or every point where a human could interfere so Diva doesn't have any white lists it doesn't have any multix no oracles no Guardians no daies is simply a way where people can either become a liquid Staker or an operator and find a balance between that um the way I like to tell the story of this design is there is the

theorem proof of stake in the beginning and then with Lio we got liquid staking um which relies on a today proof of authority model and I know that these models are constantly evolving and we're all trying to improve uh but today it's relying on a list of white listed operators with rocket pool this setup went fully permissionless but to go permissionless it means that you need to also provide a way where operations are collateralized and today that's using a model where each each mini pool or validator is run by one operator and the evolution that Diva does to this is it becomes distributed so instead of having one key per validator we have multiple nodes operating together using DVT each validator and that opens a lot of

possibilities I like to think about the AIT as a layer two for staking because it gives you all these advantages of high availability together with this ability to rewrite the rules so I like to say there's liquid staking there's economically trustless like rocket pool and there's distributed validation and when you combine these three uh that's kind of the the architecture that diva is going for so in a comparison what you will see is that on the Staker side it's very similar there's no minimum amount you can get a liquid Stak in token you do use a smart contct track that's I think common to everybody where things get a bit more interesting is on the the operator side because we map this one

validator one key and by breaking that Paradigm we managed to lower that collateral from around 10 e in total value for Rocket pool which is uh one of the lowest uh to one one e per key share and I like to think of this uh Spectrum kind of of where where we lay in terms of um accessibility so we run a little bit some numbers and what we discovered is that there's a lot more people that are today holding one to 10 e than people with 32 e and more and the potential if we manage to lower this this collateral barrier to entry is that we could be on boarding about 10 more people to staking which I think can be very powerful and I think that can help

us reverse a little bit this trend of centralization that we see in the higher GDP regions um the technology behind that is distribut the key generation using multiparty computation so in Diva every time a validator creates a key it does that in a distributed way where each node generates a key providing collateral uh and it uses a consensus of two out of three for signing um that means that it's much more resilient to failures if there's a data center or somebody goes off or there's a bug uh but it also opens this possibility to be self-healing so if we lose one key the network can detect that that that key is missing and create a new one so the

network can recover 100% liveliness so there's many advantages I'm going to go a little more in detail into what that those models look like but first the architecture looks a bit like this so you have on one side the sers and this and the smart contract uh and that creates a validator whenever it's possible and the way operators connect is basically operators keep a balance with a smart contract on coll Al and they say look I would like to collateralize up to 100 e or 10 e or 1 E will you give me keys and what that have what that does is every time a new operator can be created Keys get distributed to The sufficiently

collateralized Operators um these operators connect via peer-to-peer Network so there's no topology there's no structures or a given um kind of topology of network but instead they kind of reain evenly and that means um there's a more fair and random distribution um one interesting component as well which we won't go into detail is a zero knowledge proof rewards Oracle uh which the team is developing um and that removes one yet one more element of trust but kind of we're trying to do is is is go back to this really strict design uh that's trust minimized so I mentioned there's 16 nodes operating each single validator

the thresold is 66% that means that up to five notes can go offline at any moment without impacting operations uh if you were to slash that we require also a consensus of enough people so accidental failure has much less impact mathematically um and there's a few specific cases that we can look into so the motivation for this is because there is solar staking there is centralized there is liquid staking I think there's a lot of potential liquid staking because it allows for this specialization but but if you look today uh we're still far from reaching the the decentralization we would like and if you look today at the three top seconding pools you're still controlling over 50% so it's really important that every model that we look into is not

only TR minimized but also that has good incentives and scalability to be able to um kind of become a a major solution so the main problem that diva is trying to fix is how do you remove that trust from the node operator because when you're sticking with somebody you're essentially trusting them or yourself um and with the rocket pool model of bonding what you have is you have one single person responsible and they're posting uh collateral for that but you still have that with individual entities so once you start crisscrossing these responsibilities everything becomes a little more uh versatile but you could also say there's also bad parts to that

which is that if that becomes a lot cheaper the network attack could could paradoxically become um easier because you could have less Bond so it's very important that when you we design this risk models that we do it with the awareness that by lowering this virus we could be making attacks easier um so yeah this is something that the team has put a lot of work into into how to like um be civil resistant and how to protect against each of the possible failures so um I'm going to Showcase just a few things um that may be interesting technically this conal talk that you can see online is called the magic behind distributed validators by Prada uh CEO

of Samir who's somewhere at the event um and this is a bit more technical but the way validators are created needs to be random because there's a certain kind of attack which if you would want to control the network you could simply provide a lot of collateral and start receiving all the keys and then start signing so to prevent this the architecture in diva is using some ontain Randomness um where operators who have enough collateral are playing a kind of unchain Lottery to see who can receive those keys and when they play this Lottery that ensures a fair and completely random distribution which yeah uh is much more secure although it

can be frustrating for those of you running testnet and who are still waiting for Keys um but kind of the result of this process is that you end up with a bunch of signers who end up having keys to different different groups of validators and that Ure ensures that it's much more resistant to take over um we have this analogy as well that dvts to validation like MPC is to custody when you use NPC thr signatures you need to collect all the C signatures in order to transfer but when you use that for validation it's so much more complicated because you need to be doing that continuously all the time on time with latency impacts with different

clients behind so this is a very hard problem to solve but we also think it's very worth it because when you look at what is the impact we are going to look here just at the downtime Improvement but if you have a node which is very bad it has 5% downtime what you would have is something like this where it looks very spotty so what we're visualizing here is that we have 16 different noes and each one is down 1 12th of the time um that would be in my mind like an office where employees are constantly sick but what this team kind of team validation allows you to do is that because the threshold is much

higher when you have this group that is constantly failing the combined entity the team never fails that office never closes and it can always provide service in a timely manner and I think that's what we're after when we talk about resiliency inv validation and here's a way to visualize how increasing the DBT set size has an exponentially increasing benefit um in terms of app time obviously there's also cost to that so um there's a point where is enough that I I think that's why the 16 is a good number um and when you visualize that in terms of where up time is in the network what you can see is that this is distribution of current inactivity

percentages you can substantially improve that um beyond the level even of professional validators in a theoretical level so I think that's a very clear image there's another one which I like which is in terms of slashing protection so if you have a 33% client buag uh everything's lost because of the correlation penalty not great if you have a 10% client bu and you're affected by one of your clients then you could have a loss of up to 30% in those validators but when you're aggregating behind that diversity now has a whole other dimension which is if you're running a diverse validator set behind your DVT your loss even if you're

affected by a 20% client buag could potentially be zero and that's because you have a consensus before you go to main net so that's really interesting and again just another example um so we're e taker we um can talk about how to run a node so to install Diva for you for those of you curious to try is today running on girly so it's just a Docker package with a Docker compose you can install it um in the future you'll deposit collateral to activate the validators and then you get this additional operator rewards um we built this web UI that looks like quite friendly and the idea

is to appeal to home stakers who maybe don't have that much experience and to really give them the simplest possible version um and again the the difference is where on ethereum you would need three to e on Diva you need one mapping one key is now 16 you have the consensus protection you get the Liquid staking and you get the the the distributed generation so in a way it's like everything that ethereum staking does it Diva kind of extends uh for those of you wanting to run it it's it runs on on top of an ethereum full node so you can run your existent setup can connect it to that and the diva itself is very lightweight um and basically it manages everything for you there's no need to do

to create the keys everything's done behind um behind the client so again there's the three Journeys I think a lot of you will be interested in the run and node version so yeah if you're curious you can play with it testet um maybe yeah if you're curious about the process it looks like this in the simplest way you could provision a VPS hopefully use your own machine get clone and then there's kind of a two-step install process and you will be running um there's already some great tooling that uh somebody in the community pet is there uh where you can play see like what's active or not and that's that's quite nice to see um and what we' been doing as well for those of you who don't run notes but are active

on defi is Diva has had some pre-launch early Staker initiatives so if you have eth there's two ways to participate there's um an enzyme Vault where you can deposit either e or S and that would once Diva launches it will be unstaked from Lio or just stake directly into diva and that will also give you Diva locations so anybody who wants to support the launch somehow is a very simple way on chain or if you're running mainnet validators and you're looking to kind of transition and get some of these advantages there's also a validator migration initiative where you simply signal that you want to move you sign a transaction non custodi and later you can do it so so that's a bit what we're

doing uh you can learn everything at dak.com and a lot of us are here so yeah happy to answer any questions any questions from the audience thank you uh nice talk I wanted to know uh so usually you would like to run multiple validators in one single cluster and I was curious about uh how many you know in this terms of scaling how many validators have you tested running on the same cluster together it's like it's possible to run hundreds

of validators maybe thousands of validators in the same cluster so um we don't have an exact number or recommendation but because D.A connects to an existing ethereum node I expect that the bck would probably be in that node so I would say probably hundreds or thousands but uh it hasn't been tested extensively yet my recommendation would be that you don't put more than a th eth per machine just out of economic sense any more questions all righty well let's give another big Applause for uh Pablo