Secure Off-Chain Data Availability for Ethereum L2s with Celestia | Mary Liu | ETHTaipei 2023

Mary I'm presenting from Celestia the title of my talk today is securing securing off-chain data availability for ethereum l2s using Celestia so a little bit of an agenda today for today and this talk is actually perfectly perfectly timed after the dang starting talk so you guys are already familiar with some of the concepts um so the first thing I want to do is start by addressing some of the Core Concepts of uh of our building blocks today so that's scalability and modularity and then second I want to start talking about celestiums celestiums are ethereum l2s that use Celestia for off-chain data availability and then finally I'll go into some other aspects of our ecosystem including unique aspects and potential that our ecosystem empowers AKA Sovereign roll-ups so let's get started so this is the screenshot that I took last night from both etherscan the ethereum gas tracker as well as L2 fees so on the ethereum um on the ethereum gas tracker uh you can see that uh gas fees uh even in a somewhat of a bear Market are still quite expensive so you have an average of two dollars uh per transaction and for example a Open Sea sale costs about seven dollars uh as of last night people have attempted to scale this and lower transaction fees by batching together uh transactions into a single transaction called Roll-Ups so you see the performance of Roll-Ups on uh the left side of the screen um so these costs range from Seven cents to about 46 cents um and that's still too expensive because uh Roll-Ups still require that transactions be processed on the ethereum mainnet and rollups still have scaling limitations of their own so in the next Bull Run uh when you see you know other Deeds being minted for thousands of dollars again you can expect that these transaction fees will rise as demand for ethereum Block space Also increases and this of course results in exponential gas fees so here's a quote from a dank rod on a recent ethereum Forum post uh just like the base layer the price of data for Roll-Ups will be volatile and based on demand if there is a massive Bull Run or a big crisis a lot more transactions tend to be made and Roll-Ups will not be isolated from uh the the high intensity of transactions so rollups will have to be able to cope with volatility in any case um next our second core concept for today modular blockchains modular blockchains decouple the core functions of the blockchain so the core functions of the blockchain are execution settlement and data availability now you can kind of see the progression of software from the monolithic era era to the modular software era to finally the modular blockchain era you can also think of this progression in the progression from Bitcoin to ethereum to Celestia so the Bitcoin Network uh the Bitcoin Network enabled secure and Anonymous peer-to-peer transactions designed to be a simple Ledger of transactions ethereum was the first smart contracts platform that enabled programmable money that introduced more complex logic without the need for any intermediaries or counterparties it's also the reason we're all here today uh and then finally Celestia so Celestia was an Innovative New Concept in the idea of decoupling the core functions of any given blockchain so instead of a monolithic blockchain performing all those functions and really kind of falling into the scalability trilemma of uh increased costs in return for increased security uh Celestia decouples the blockchain so only focuses on data availability and consensus because it solely focuses on data availability it's designed to be highly scalable and optimized for ordering data and making it available now how does it work so you see monolithic blockchains uh to your left in monolithic blockchains if such as ethereum ethereum full notes have to download all the data in a block to increase block space you also have to increase Hardware requirements modular blockchains on the other hand rely primarily on light notes light notes using light nodes you only have to rely on multiple rounds of downloading smaller chunks of block data so you can see a smaller chunk of block data being downloaded and then because there's multiple rounds of uh of verifying that the data is available and published you increase and hit that confidence threshold that's predetermined so you can receive a larger amount of block space in return for a minimal cost increase we also have documentation on this you can essentially run a Celestial light node from your smartphone so here's a map of uh of our validators as well as light notes you can see that it's truly decentralized right there's light nodes all over the planet uh and the light note the number of light nodes scales as uh as the Celestia Network increases in use as well uh so the differences between like clients in monolithic blockchains as well as modular light notes like the celestial light node so in basic like clients basic like clients simply verify that the majority of validators have approved a block so they only check the header uh so in that way like clients are fairly unsecure modular like clients on the other hand um I both check the header and then also download a portion of that data so if the data in that block is actually invalid then it'll know thus like the celestial Life notes have the same security have similar security guarantees as um as monolith blockchain full nodes so now we get to celestiums celestiums are an L2 chain that uses Celestia for data availability but uses ethereum for settlement and dispute resolution for a handier visual you can see the uh the infrastructure stack behind me um so on the left hand side you see celestiums relying on ethereum for settlement and consensus and using Celestia for data availability now how does it work so this is a high level architecture of uh of how ethereum l2s can leverage Celestia for data availability so ethereum L2 operators post-transaction data to Celestia's data availability layer um there it's placed by blocks uh there it's placed into Celestia blocks by Celestia's proof of stake validators data is then relayed in the form of a data availability attestation from Celestia to ethereum the attestation is a Merkle root of the L2 data signed by Celestia validators attesting to the fact that the data is in fact available on Celestia now the D.A Bridge contract over here is a contract that sits on ethereum the D.A Bridge contract called the quantum gravity Bridge is slushy's data availability Bridge contract that verifies the signatures of The Da attestation from Celestia now when the L2 contract on ethereum updates its state it checks to make sure that the correct data was made available on Celestia by querying into the da Bridge contract instead of relying on transaction data being posted as call data to ethereum which is much more expensive and difficult to scale now the bridge contract returns a positive response for any valid attestation that has been previously relayed to it and negative if else so here you can kind of think of uh of a celestium as a data availability committee with extra economic guarantees on data availability because the committee can be slashed if it misbehaves um this is possible because lightnodes on Celestia's Network can detect unavailable blocks with data availability sampling which we've discussed previously because of data availability sampling Celestial light nodes share the similar security guarantees of an ethereum full node and thus can automatically halt if the validator set becomes malicious Celestia relies on tenement with with proof of stake so Celestia is a proof of stake Cosmos chain that relies on tenorment um and has a larger permissionless group of stakers on the network thus once again D.A attestations that are incorrect and invalid can be penalized with slashing so some comparisons with uh dank sharding which you guys are all familiar with thanks to the presentation before um so first in our comparison chart we start with block times um block times are more or less uh uh negligible between a couple of different da Solutions you see Celestial block times are approximately 15 seconds um and ethereum with the EIP 4844 are about 12 seconds I'll focus in on eip4844 you can read the other two comparisons yourself as well and then um and then what matters more is kind of the time to finality uh so between ethereum and Celestia the times of finality even with the 4844 is 12 minutes so ethereum's block production engine titled ghost um is uh uh relies on probabilistic finality and uses Casper to provide faster finality so even with faster finality a faster economic a faster guarantee of economic finality the Vlog time is still about 12 minutes Celestia uses tenorment for consensus which has single slot finality so once a block passes Celestia's consensus it's finalized thus uh the time to finality Force last year is approximately 15 seconds uh next data availability sampling so ethereum with eip4844 as the previous presentation mentioned does not support data availability sampling and likely won't for the next two to three years um but Celestia does support data availability sampling it's ready now uh we're going to mainnet soon um so developers will be able to use Celestia's Network which has data availability sampling to verify data availability on ethereum users must still run full nodes and download all the data so you see monolithic blockchains as transaction number increases so does the fees uh and while cheaper with 4844 fees will still increase as well but for modular blockchains as mentioned previously before as the number of light nodes increases in the network um the celestial Network scales with demand so transaction fees stay fairly low now let's talk about like lines so we've hit on this previously um Celestial like clients uh rely Celestial light nodes rely on an honest minority because of data availability sampling like clients can't detect if an invalid block is produced by a dishonest majority of block producers because it relies on validators vouching for a block nodes with data availability sampling get an upgrade in security because they can verify if invalid blocks are produced so that this is only it happens if the da layer focuses only on consensus and data availability and then next a little more on uh in ratio coding and the encoding proof scheme so Erasure coding is an important mechanic that makes data availability sampling possible this is because it extends a block by producing producing additional copies of the block data this additional data creates redundancy and gives stronger Securities guarantees for the sampling process nodes still however may try to incorrectly encode data to disrupt the network so you can see here ethereum with EIP 4844 eigenlayer as well as polygon Avail all rely on a type of validity proof uh so these kind of work in the same sense as validity validity proofs work in ZK roll-ups so each time a block is produced validators must produce commitments to the data which nodes verify using a kzg proof this provides that this proves that the block was encoded correctly however producing commitments for kcg proofs producing uh Kate commitments for kzg proofs requires much more computational overhead which means it increases the cost and increases the amount of Hardware that's required generating commitments however doesn't carry as much overhead when blocks are small as the blocks get larger commitments for kzg proofs will carry a much higher burden to generate no types that are responsible for generating kcg commitments will require much higher Hardware commitments Celestia is unique because it uses a fraud scheme to detect incorrectly encoded blocks Celestial nodes don't need to check if a block is correctly encoded they assume it's correct by default so the benefit is that block producers don't need to do expensive work to produce commitments for the Erasure coding and then finally I'll talk about some other unique aspects of the Celestia Network so regarding Sovereign Roll-Ups Sovereign Roll-Ups developers can Define themselves the execution and settlement layers and then use Celestia for consensus and data availability now what are some benefits of sovereign roll-ups with Sovereign Roll-Ups you inherit that same scalability I discussed before as demand for Celestia's block space increases Blocks place block space increases as well and the light nodes increase alongside that so scalability is received through Celestia's data availability sampling technique and then Roll-Ups that deploy in Celestia share security via Celestia's validator set these Roll-Ups also maintain sovereignty because they Define both the execution and the settlement layers themselves so previously when folks wanted to launch their own monolithic blockchain right we can think of a lot of different alternative layer ones they had to build their own security from scratch they had to build their own validator set with modular blockchains you inherit security from day one so you can deploy your own rollup Define your own execution and settlement layers and then inherit security through Celestia's decentralized validator set so there's a lot of different configurations that allows for hyper customizable customizability as well as versatility for the end developer so developers can choose evm Salon a virtual machine environments uh you can choose to deploy a move roll up you can choose to either be part of a shared sequencer net Network or rely on a decentralized sequencer the possibilities are kind of endless with Celestia right now a number of Roll-Ups as the service projects are in the works and work and work with projects such as D5 protocols and nft projects to deploy Roll-Ups in a customized fashion so I've included two links here one to our build modular documentation as well as another to uh the GitHub link for the quantum gravity Bridge which is the data availability Bridge contract on Celestia please feel free to contact me either through Twitter or Telegram and then please come find me if you have questions as well thank you thank you Mary to bring this wonderful talk so now we are open to two questions foreign so thanks for the amazing talk so I got one question so how secure is the Celestia consensus Network uh the consensus network is tenorment so tendermint has never been hacked um it's the tenement bft consensus okay so uh uh I assume it also depends on how many stakers in the network right is it like yeah how many stickers in the network yeah how many stickers in um because the consensus depends on like the robustness of the stickers If the sticker is more and that means the consensus algorithm will be more robust right yeah so I I just want to ask um yeah basically the security yeah so security increases right with the number of validators in the network so right now we have 100 validators participating um and then post launch you know we expect more validators to participate okay thank you very yeah this is will yeah and I I'm curious about why you put a eigenator uh to in the combination comparison comparison table Yeah because I think like uh Celestia is a is a layer one and eigen layer is something like a protocol and so if you talk about the finality of eigen layer so it means it will be the same answer with uh you're Syria because right now they build in ethereum so if today the building on like tender mean or Cosmos chain the finality will be the performance of the cosmos chain so I'm just curious about it and also give you some feedback about it yeah I mean no that's entirely correct so eigenlayer is a set of smart contracts right on ethereum so it does share the same finality as ethereum I I I we like create did that chart um just because there are a lot of comparisons that we wanted to address um because one of eigenlayer's products is eigen da um so in that sense it's just a comparison of data availability Solutions I did I know from risk Finance um I've um two quick questions one of them like when you're taking the the smaller blocks to do the validation on that um how how are those blocks assigned are they assigned randomly and if so what pseudo-random number generator do you use for that um and then the other one it was 15 seconds to get to final because of the Assumption of the block being valid but if there were to be a block that had a fraud in it um how long would the network take to react to detect that and pull the block on or can you describe that one because I assume that would take extra time thank you uh are you I'm sorry are you referring to like data withholding or all right like what could you just restate yeah certainly um so it was 15 seconds to for the um for the block to be finalized um now that is based on it's faster because of the the fraudproof assumption right so there doesn't need to be fully validated because you're waiting for validators to come back with like a counter example maybe something like that but how long does that take to happen if there is a counter example that needs to be checked and validated and then the slashing decisions need to be taken and does that add extra time to the block processing time yeah that's a good question um I don't know the exact number of minutes time I I do know recall off the top of my head right now sorry I'll add me on Twitter or Telegram and then I'll I'll get back to you okay so if there are no question then we will finish our day one morning session our after recession will start at 1pm so yeah if you are interested in some of the talk please come back to the building and we provide a co-working space at the at the building B which is on your right hand side and we also provide the Boost some of the Boost so you can uh go and visit if you are free after lunch it's just close to our next building thank you foreign