Ensuring Data Availability in L2s by Vincenzo Furcillo | Devcon SEA
Devcon 7 · 26 Nov 2024 · 8:43
Vincenzo Furcillo examines data availability in Layer 2s at Devcon SEA, covering why rollups need a reliable way to publish the data behind their transactions so anyone can verify and reconstruct state. The talk surveys approaches to keeping that data available and affordable as Ethereum's rollup ecosystem scales.
Speakers
Topics
From Devcon 7, hosted by Ethereum Foundation.
Transcript
[Music] [Music] hello hi everyone my name is Vincenzo I work in research at l2b and today the talk is about data availability risks and how to make a d secure looking at it through the lens of The l2b Da risk framework so to start data availability is crucial for L2 security it's um nodes need data to reconstruct the state and Advance it and this is important for both kind of constructions both for ZK constructions
or optimistic constructions and the reasons differ slightly for example for an optimistic collup you may have a malicious State root posted to ethereum and you need the data to compute the the fraud proof and invalidated while for ZK constructions you may have a correct State being posted to ethereum but um without the data to prove the ownership of your individual assets uh your funds may be Frozen and if you take into consideration also the possibility that the malicious operator uh may ask for a Rome to to make the data Available to You funds
could be stolen also in in the ZK case and so since the risks are critical how do we make a daer secure and the first thing to consider is economic security which is definitely not a meme and it is important for liveness so to make sure that the data can enter The Ledger so there is no sensorship on the D layer and also for safety so uh that once the data enters the Ledger is not reor away from a for a deep reor and this is the current state of the D layers listed on l2b and you can see that there are some project that have sted assets but slashing
for non-attributable faults like data availability withholding attack for example is very hard in uh in Celestia for example you would need a social Fork which is expensive and hard to pull off for uh aen you would need a a token Fork but the takeaway is that most slashing mechanisms are still untested and some are still under development like the aen one so how do you protect against the most crucial kind of attack in data availability the data withholding attack you need data avilability sampling and this is in production for a couple of projects like Celestia and ail but the
Assumption of Thea availability sampling is that there is a minimum number of light nodes on the network that are sampling this data and this is very hard to verify so if if you're working on it uh come talk to us and another thing that is very hard is the block reconstruction protocol so once these light nodes sample all the data um you want them to be able to uh communicate peer-to-peer and reconstruct the full block so data is is not lost and this is still under development for most projects today but the DL layer is only one side of of the equation a secure D layer what's also important is to have a
secure da bridge and so for an external da that um that is may be used by an optimium or a validium you need ethereum to receive a proof that the data was made available on this external da layer and this proof then needs to be relayed to a Smart contract on ethereum and this smart contct is called a DA bridge and without Bridge ethereum has no proof of the availability so it is very important and the bridge needs to be integrated in the proof system of the L2 um today there is no uh LDA L2 that integrates with a DA Bridge which is pretty bad but hopefully that will
improve soon so briefly looking at what makes a DA Bridge secure cuz we looked at the D layer but what makes a d Bridge secure is three main things you need to look at the community community security so who provides the attestation to this bridge is it a small committee is a big one is it your friends is it a validator set then you need to look at relayer failure so who can relay the da attestations and what happens if it goes offline or turns malicious and also Bridge up curability so are these attestations immutable or can someone change the history commitments of these
da attestations and you can see from the red on this slide uh also there is a lot of work to do on on the bridges and if you're wondering about Dax because so far we have seen external da layers but there are also uh data availability committees and there are many of them some are better than others but you can get the the full list on the L2 be website and if you want to find out the more more details about this uh risk framework you can find it on the l2b Forum and you can scan the QR code to to get to the link and since I have some time left maybe I can say that um D layers as you have seen are have different
characteristics different risks so despite da being cheap and some people saying it's a DA is a commodity I think it's important to to look at these differences and actually take seriously the the risk that may may come up from from using external Das and that's it for today Applause please so it is more than an important and thank you for summing this up so nicely in like in short minutes of course we could talk about this for and we need to talk right so uh please questions we have two minutes for questions where do I throw this maybe you can just um say from from your
perspective from the background that you have like in the risk management and these kind of things how do you see how do you see the state of of of of of this today right yeah this projects are pretty new let's say they they launch may not maximum a couple of years ago so there's still work to do and teams are are serious so they're working on it and right now we have the risk frame work to kind of get the state of how things are at the moment and not there is no project that is all green CU they just started let's say I mean they've been working on it for a long time but um Mains are more recent uh but I'm sure
they they'll you have get better soon yeah you have F of course I that's good to hear H did anybody thought about any question no yes please raise your hands yep please please please use the microphone so the people on internet can hear as well yeah so uh you know you said that I da use is a DA services and others are public blockchains so could you give more comparisons of these different kinds of Das yeah in this slide I showed uh D layers so we divided in two sections on on L2 be we have
public data availability providers and custom so in the public ones we have public blockchains like Celestia Veil um and also agda as a DA service because it's a it's a public layer that is meant to support many projects using it while on the other section you would have uh data availabilities committees um that usually are deeply integrated with one single L2 okay thanks for the question and for the answer thank you Veno again and thank El to be to everything that you guys are doing thank you thank you thank you guys so Applause please for venzo yeah