DAPPCON 2023: Gnosis as an L2 - Dapplion

Transcript

what a wonderful week to be here so many beautiful faces and Friends hello everyone so today I'm going to speak about the nosis chain and I want to do one question before how many people knew be please be honest that nosis had a chain raise your hands very good very good it's it's surprising how many people sometimes don't know this but cool so no chain it's a wonderful project that I'm passionate about I'm the for coordinator in CIP and today I'm going to focus on how can we improve the security of the chain and especially the bridge using this wonderful technology that has been coming from the L2

wall so just some background nosi started as XD being a POA connected to ethereum with a native Bridge we have added many more since then and since December we have done the merge very excited now we have a permission L validator said so the security of the Native value is pretty good but the security assumptions today for the value that comes from ethereum still depend on Multi Bridges it's important to note that even if the permissionless set of the honest validators gets corrupted because the bridge operators run independent software they will still be protected but if you corrupt the bridge operators

bad things can happen so on the far end of the spectrum the end game of rollups like this is not what we have today unfortunately most of them have multi and fast upgrade mechanisms but let's imagine a happy future that's what you get with rups you get a situation where even if everyone on the rupite is corrupt and you can only trust one single party exist somewhere in the planet assisting the bridge you will always be able to get your money out no matter what this is awesome and why ethereum is going so hard into rollers right now so let's explore what will happen if we try to move a little closer

there so I love this table from matter laabs it's a bit old but very Illuminating we can see here how side chains if you can corrupt the brid set you can steal fund you can do bad things with rollups you cannot with optimistic you have this nasty with draw delay but something we can work with and so exploring the middle ground between these two there is this very interesting solution called the validium where you sacrifice security by posting data somewhere else so if you can corrupt the data abity committee you can do bad things not as bad things but still put the state the network in a state that is

corrupted and only the attackers can undo these things so they can do Ransom and a lot of bad stuff so in nosis we are here multi like many other parties and can we do better could we ideally go to a place where we have these wonderful non-existing security assumptions where you only trust that somewhere someone is assisting the bridge and you can still get your money out well yeah L2 it's all about Bridge engineering and it always has been so Martin goelman proposed this construction some time ago to me and I've been exploring I think it's fascinating so I'm going to share it

with you what if we had a validium that fall back into Z rollup Security in case of a problem so let's explore that we have the nosis chain the beon chain as today validators are doing their thing they are doing blocks they are attesting they are finalizing the blocks and you take this finalized checkpoint and you throw it to L1 you have a consensus proof that shows that this is what 2/3 of the validator said agree is correct and a validity proof that shows that the state transition was done properly only then the submission is accepted to the L1 Bridge contract at that point a challenger window opens and someone can come and claim and dispute

that this checkpoint attest to data that was not made available if that happens a dispute period opens and the challenge can either fail if someone say the vators post all of this data on chain proving to L1 that indeed this checkpoint was available or the Challenger the challenge succeeds with no one doing anything for the window and then this checkpoint is reverted the cool thing with this construction is that you will never ever have a situation where a checkpoint goes on chain that is either incorrect uncanonical or unavailable so you get around this like nasty situation with

the bidum where they can corrupt the chain that's not going to happen to the cost of two things a dispute window which affects ux significantly and this assumption that someone will challenge so this construction of optimistic availability in theory should not work this is why plasma failed at its time so why is this specific challenge sound first I want you to explain why data availability is hard the problem with data availability is that it's a non uniquely attributable fault and to explain why let's go quickly with this famous example from vitalic so we have case one where the proposer is

malicious it publishes data that is not available so the Challenger has to say hey not available but then immediately after the Challenger the proposer goes and publishes remaining data that's case one and case two the publisher is honest but the Challenger is dishonest and with fully available data it also raises the alarm the problem here is that for an observer at time three you have no way of differentiating these two cases and this matters because this Observer is the bridge so you cannot go to the bridge and say who of these two is the malicious party so because you cannot rute fault this situation is not

economically stable if we had a case where the Challenger had the expectation to make money out of a challenge it could just challenge all the time that like there is no tomorrow and money pump if it had the expectation of a stable economic it could just post challenges all the time forcing everyone into downloading all the data and having um a bad Network it only works economically when the Challenger loses the problem here is that you need altruistic Behavior you need a challenger that wants to challenge for some reason even though it is going to

lose so why in our construction why the hell would the challenge the Challenger lose money well because in order for this challenge to be even possible you need to have a mission of something that is unavailable and finalized what means that not the whole but 2/3 of the nosis beon chain set got corrupted malicious or something really bad that is both very rare and something really worrying so if that was to happen there is a strong motivation for our Challenger to come and save the chain because we would do something like social slashing or some mechanism some strong fork in order

to recover control of the fork but this would be both a rare event and also time bounded so this is what makes the challenge stable so now extending like Neo new column with this thing the penalty is that yes we have a withdrawal period but we don't have a situation where if the chain gets corrupted you can steal funds so I think this is very interesting we will explore it but it's not only this we can do so much more like starting from a multi seek we could throw a thow prover and now even though we had a withd drawability delay The Operators

could never steal fans which is much much better if some team finally cracks the ZK evm type one um price we could throw it there and then we would have no withd drawability delay and the assurance that the bridge operators could not steal fans with this it would be even better because we would have both a permissionless set not pay data and not have a situation where a malicious set could could steal the funds so to to explore this idea further and I'm stealing the questions from you or you can just write this down and ask later why does this whole thing need the

consensus proof well this is important because otherwise the challenge would not be uh a unique event so the fact that it has to be finalized first reduces the the possibility of having this delay for the validity let me let me see here so for the validity proof the corruptor could steal the fans but here we would need them to have some fraud prooof construction so then the system is not as good but it also makes sense so for the consensus proof we could then opt out of having to prove the whole Pro

system because this is very very difficult like putting the whole bigon chain consensus the Casper finality gget in a proof is proven to be the very difficult there are teams working on this so we could have something first like some smaller DIC the problem is that having this intermediate layer would pretty much break the security guarantees that we want to have eventually and it would mean that we have to rely on a permission set again so this is something that we could have as a B1 and then as an end game go for the Full Construction with the proof of stake circuit construction another question for el2 engineers is okay great this is side chain but is the state actually derive from L1 or no well in normal operations

no this ien would work autonomously and they would decide how to pack the blocks because data is not meant to be posted on chain unless that there is active sensorship so to ensure that even if the Bator set gets corrupted they can exit we would have some mechanism like most ORS do where there are Force inclusion transactions that you can submit to the bridge contract and if the bridge contract doesn't see this on the blocks that you post later then some emergency mode would be activated and that's what allows you to unwind all your positions and exit from from the side chain SL

rollup so is the emergency mode recoverable I would say no we are researching if that could be possible but I think in this construction it will not because if we have to have a situation where the validator set got corrupted it doesn't make sense to basically return power to them so there has to be some intervention to change the consensus verifier circuit into accepting a new Fork so because of all of this something that would probably have to happen that is an important J consideration is that now all the L2 block producers would have to track L1 in order to include these transactions I think we can

optimize this away but it's another U thing to consider so so this is basically the end game I was talking about but I want to give a huge shout out to the Hashi team from nosis where they are building this bridge implementation that would act basic as a multiplex into different types of bridges so what we could have today with current technology is the krent native Bridge with Omni bridge and a Celestia Bridge together with the AL Cent so we would have the best of being safe against implementation bugs also having approve of the consensus with the alter like client it's not as strong as what we have we will have eventually with the

full beon chain but is still much better than not having anything so what this would give us immunity against um correlated like uncorrelated bus consensus takeover and I think they would we would have this ready I don't know how soon but it's very exciting and all of these constructions I'm talking about would probably be plucked into Hashi with different models so yeah to summarize the idea of the validum with Z rollup we would remove all assumptions at the cost of having a withdra ability delay and this well capitalized challenger but yeah very excited very bullish thank you so

much question time I don't want to ask a question so we're there just one second that they bring you the mic hi um so uh I was just curious the price of tracking L1 is like at least double if not more compared um right nosis chain to uh ethereum like how are you going to convince the validator set that it's okay to increase their you know cost by 3x um for a maybe Improvement like what's the like what's the Thought there yeah so I think you're talking about this potential requirement

that L2 block producers will have to track L1 yeah exactly so really the the requirement here is that the deposit transactions will have to be included in L1 blocks within some time frame that would probably be like 12 to 24 hours so something we could do at the start is make sure that there is some entities that commit to including these transactions and as long as someone does it then we are good so not everyone needs to just some or so everyone doing it gives you the maximum security ass because if no one is doing it uh then the the like the chain SL rollup will fall into emergency mode and basically kill lifeless so you want to make sure that that doesn't happen unless there is

an accident but you can try to have like an maybe a happy mode where if these parties are doing it then no one else has to do it but there is still some mechanism that people will just step in and include these transactions if no one else is doing it awesome thank you sure any other questions no no questions now is the chance eh no okay well thank you for coming and thank you thank you so much