'Defining the MEV Game on L2' - Georgios Konstantopoulos

um tina do you want to introduce the next speaker yes um next will be georgios is going on you hear me yeah we can hear you okay wait and now you can probably also see my screen um cool is the screen on too yes okay fantastic so thank you all for coming thank you carl for the very energetic presentation as usual my presentation is going to be a lot more vanilla and we'll make a lot of assumptions about what the audience already knows because we've already heard a lot so skipping through i'm georgios i work at paradigm venture capital firm and i'm a research partner here so let's talk about mv and before we talk about mev in layer 2 maybe it's worth kind of understanding what the questions we should ask when talking about mvvr and the firstly let's preface this by saying that i don't think that we can remove remove it fully but we can minimize it and there's various layers of the system which we can touch on towards minimizing the movie one of that is on the protocol layer on the very base layer there's a lot of parameters there's how block producers are chosen because maybe the block producers as we saw earlier in the discussion are random but maybe they're predictable and if they're predictable what are the implications of that um for mbv maybe if many blockbuster users are predictable maybe it's more likely that they collude and that they end up maybe doing multi-block mvv extractions like time-banded attacks it's also another good question to ask is are transactions in the clear so are transactions submitted to the sequencer encrypted or are they submitted clear text if there is if they're encrypted um what is the mechanism to decrypt them and so on with respect to ordering who owns the ordering is it always the sequencer or the block reducer or the miner that owns the ordering operation or is there some consensus and forced operation which randomizes transaction ordering this is also very important um what about finality does the protocol have finality how long is that finality finality is a big factor when you're thinking about time-banded attacks again in the multi-block mv extraction scenario separately from that on the network layer there is a question to ask around byzantine fault tolerant proof-of-stake versus the proof of stake that we have in ethereum for example in tendermint um the mav cannot there's no notion of time-banded attacks because once a block is mined once a block is validated there's no way to roll it back but perhaps the mv extraction happens at the network layer during when the gossip protocol plays itself out and validators try to decide which transactions they should mine on the network layer also because i didn't add it because we've talked about it a lot and after all we're in flashbots mev rose there is flashbots also which allows you to basically instead of having the main communication channel you try to introduce another one such as the the flashbots uh kind of like second mempool and that allows you to introduce new rules on the network layer there finally on the application layer there's things which you should consider around mev such as you know thing thinking about it from first principles mv is exposed when anybody can when there is like race on the application layer it gets exposed when there's races towards touching the same piece of state for example when there's um let's say an order on uni-swap and they both touch the same the all orders that's the same reserve variables so something that could improve things or for example in liquidation so something that could improve things would be that maybe you try to introduce some ordering on the actors that try to go after an mev opportunity and an example of that would be keeper dao's k compound protocol where instead of allowing anybody to liquidate you know they just allow certain parties to execute liquidations based on the dow elected liquidators similar similarly done by um by b protocol again on liquidations so like the giant notions that you try to remove this sort of anybody can liquidate anybody can herb and try to introduce different structures in the way that actors interact with the system so now that we have kind of a basic questions on how we should be approaching the problem it's worth maybe looking at it from the angle of layer twos so just to say the very basic things the layer twos like since the dawn of layer twos with the side chains papers what they were saying is that they let us experiment without fear of the layer one like weird things happening to layer one so we can start introducing crazy features on the layer two while being relatively conservative on the layer one but the kind of layer twos that we have right now which are mostly rollups whether it's the optimistic or the cta variants more or less the mental model should be that it's just another layer one again with respect to mev and the previous questions and that it just has some additional mechanism which allows it to inherit ethereum security beyond that there is nothing too specific to the layer two and something to note is that when whenever the layer two is more programmatic for example fuel is a much is a slightly more restricted computationally roll-up compared to let's say starkwear or z casing or optimism and arbitron kind of uh if you think about it in a spec in a spectrum and so more programmability always will mean more mev so that's how you it should be thought about on the m layer two layer and then something which was also mentioned earlier earlier is that when mev gets extracted on layer two that means it does not get extracted on layer one so in a sense we're not really removing mev by introducing layer two systems rather we're just moving it from one place to another and we introduce certain rules there and those rules may either be you know memple things the block producer election the things that carl just talked to us about or transactional randomization or who knows what else and kind of zooming in on that i'm on the mv auction i have like slightly more opinions and carl carl and i may disagree on this that in a way the mv auction forces everything to be on the sequencer and the rational sequencer probably would want to extract as much mv as they want or as they can and it remains to be seen if kind of operating this benevolent sequencer model is good enough there's another approach that we've seen by arbitrom which says that instead of kind of having like one sequencer that decides on the ordering you introduce a bft protocol on top and you have all the parties in that bfd protocol agree on the transaction ordering again that maybe works maybe doesn't for reasons field for reasons that phil has talked about in his blog post and in his talk and that is because how do you elect these people what happens does this still then does that move again the mvv from the actual consensus layer to them networking layer before when the parties are trying to figure out who should what transaction ordering should be followed so you know the takeaway basically is that there's no free lunch in any kind of mitigation that you introduce and other approaches of course are classic cryptography based techniques for kind of separating execution from ordering which only alluded to some of them with and we've had these discussions in the past in the previous rounds something to note about mev and the optimistic roll-up is specifically there's a statement which goes like because miners can certainly fraud proof for zero cost and by censoring a fraud proof um they can get paid counter factually after fraud has has been executed um you can say that all the funds in the optimistic roller could be considered mev because well the sequencer could go loot with the miners or the miners could be a sequencer and they would make an invalid state transition and you know everything would be censored for the entire dispute period so if you're thinking about this completely in a vacuum you know you're just changing sequencer miners that's it sure i'll take it but in crypto there's lots of complex interactions and we will see why this attack kind of collapses and the reason is quite simple we should think about it from two cases one where miners just do a 51 coordination and they orphan any blocks which include the fraud proof so this would be a chain split scenario where some miners decide to create them to accept the fraud proof some others don't and effectively you have two ethereums this ain't gonna happen miners are long if any kind of attack on ethereum in that way directly destroys destroys their investment so that aside there's also another approach which would be that you know maybe all the miners somehow get convinced to put their hashtag and say this transaction is bad we should ignore it um or this transaction would rather we got bribed by this party to censor this transaction so i don't have a formal proof for that or you know some kind of intuitive game theoretical argument but i don't i don't see and i don't see this possible as possible at all it destroys ethereum's base settlement assurances and it seemed like a complete non-starter to me so there is a co there is a very like small nuance here that you know we just talked about transactions which would be good but get censored by miners but what about transactions which are let's say bad for example somebody doing let's say money laundering on optimism and they want to withdraw these funds so should miners accept a transaction that um that allows the money under a money launderer to exit their funds um and on that i have no opinion ideally if the base layer is properly um distributed this should never occur like the protocol behavior that is not able to distinguish here but who knows what will happen right like our understanding of blockchains evolves a lot over time and just to kind of hint at things because i know that we have a talk in the next movie roast about cross chain mev but just to kind of drive the point home it's not just that the simplest systems and the most complex systems have various degrees of mev and of course with systems with values with variant kind of degrees of composibility or interoperability but any time any of these systems tries to talk with each other it exposes mvv and anytime a message is related from one place to another there is a party that does this relaying the e-layer you know you call it whatever you want and this re-layer will have the ability to insert their own transactions delay inclusion of your message maybe reorder the transactions and so on so the question i posed to the audience is like as the as blockchain start to become more complex and the interactions between them go up will we be able to keep up with this complexity of mbva extraction and what how will our tools evolve on all this and that was it thank you very much and i'm happy to answer any questions if any thank you georges um great talk any questions again great presentation no questions from eve tv chat um oh alex has a question um yeah but justin if you want to ask yours first if you have yeah i always have one question in my back pocket just in case no one has questions i um i'm happy to go after you um alex go ahead all right um so i'm still wrapping my head around layer two so thank you georges that was a great presentation so i have a question related to layer two submitting the roll-up batches that were with transactions within that layer two already sequenced within that roller batch two uh the data availability layer so ethereum or to a shard in ethereum and if someone is doing cross layer two arbitrage would they would they be concerned about which role a batch is is processed first by the data availability layer or does that not matter at all in the in the sense where you're you care about the sequencing of which batch is submitted first between two different layer twos does that make sense at all to you or um am i am i missing missing something here so you're saying that there's two layer twos two sequencers two data availability layers each one of the sequencers submit um the data availability batch to their own system and i am a party doing cross-chain arbitrage and the question so it's nearly that it's uh two sequences two layer twos but it's same data availability layer right so it doesn't matter i mean maybe it does matter depending on the the protocol implementation but it shouldn't matter like from a client-side perspective all you need to know is that like the the data has been published somewhere so for you like as the dex trader or whatever medium small size um it doesn't matter like what matters that you have like the guarantee that the data is there um yeah so from the perspective of um of the individual protocol i don't think that you need to care much about what happens on the remote layer okay cool there's a question from the from the chat about do you think infuria alchemy etc are going to start extracting mev soon so firstly infure and alchemy are not miners or validators to the best of my knowledge they are just node providers so they don't have any specific roles in the system or the ability to reorder transactions in whatever way whether they would do there's maybe like a dark scenario in which you know if you just start censoring transactions or start copying copy copying traders that use it um that's one possibility but in reality all the kind of legit traders they use their own nodes and they have you know like proper peering with peers that are well peered and so on so no definitely not okay so it looks like we have two more questions question from youtube do zero knowledge proofs affect mev extractability at all if you bought the argument that i made earlier that the optimistic roll-up is not affected by the attack that i mentioned uh from miners then in practice i don't think that player two and zero knowledge has kind of like any added benefit over optimistic class um plus mv so i don't think so i don't think that uh zero knowledge groups affect anyone unless you know unless of course you mean that you introduce your knowledge proof to maybe do the secret leader election or to do some things in the vdf and so on okay fantastic and i think we have just one minute um and also i asked my question which is it seems that paradigm is going big on mov like hiring a team internally that specializes in that but also investing heavily in the broad muv ecosystem can you talk about you know the investment thesis around hiring all the mev giga brains i don't know about like us hiring any of the mv gigabrains we're very excited to be working with flashbots of course um i think that mev is like an important trend um that we've we identified early on both like charlie and i were like looking at this from a while back so i think that mev is like one of the most important topics for ethereum's future and of course we care a lot about ethereum so um so it seems very reasonable to me that we'd care about mev and try to be close to it i see so it's kind of like a defensive investment kind of a hedge um you know you have all this eve you want to defend it because you know the mev is kind of a systemic risk i don't think it's necessarily a stemic risk or well maybe the nice way to put it is that we want to try or like i would want it to be a force that makes us improve our systems and make them more stable so it's uh you can call it defensive but maybe there's like a positive angle to it that's all i'm saying okay fantastic i think we're up on time and we have um our panel next with uh i think georgia is still on panel barry and stanley and david