Justin Drake 🛸 Weak Censorship Resistance

[Music] okay so back story on the slides is I've I tried to do a new set of slides for every talk and I've had five talks this is my fifth one and I was like okay I'm just going to reuse slides for the last one and then I looked back on my old slides uh from I know 18 months ago and the state of the art has changed so much that I had to do new slides from scratch so now this is my fifth new slide uh deck from from scratch um so I'm going to be talking about uh we we week sens weak censorship um and it's going to be in three parts one I'm going to talk about definitions of sensorship and weak censorship and then talk about the problems with weak censorship and then finally talk about Solutions uh to uh weak censorship so my uh you know definition of censorship is to invoke this test which I call the append test it's a very simple test if you have a transaction which is publicly available but didn't make it in a specific block you need to ask yourself could it have been appended at the very end of the block and if the answer is yes it was appendable then the transaction was censored if the answer is no then that's fine the transaction was was not censored in that specific block and there's multiple reasons why a transaction might not be appendable is because you know it might have been invalid or maybe the block was full right if the block is completely full you can't append more transactions or if the the valid data was offline and the slot is empty well there's no block so you can't append the transaction so that's my my my append test for sensorship now there's two types of uh censorship there's weak censorship and strong censorship um weak censorship is when the censorship is temporary so the transaction actually makes it onchain but it's delayed it could be a delay of a few seconds or a few minutes generally um and then there's this other notion called strong censorship where the a transaction just never makes it onchain it's kind of this permanent censorship and this is really bad because it's effectively a 51% attack on ethereum and we have some really really big issues if permanent censorship uh happens and so this talk is going to be about weak censorship which is the censorship we've seen so far on etherium we've never seen strong censorship so um yeah we're all familiar with the meev pipeline which starts with the user and ends with the attester and there's kind of this this travel right uh from intent to proposal and the reason why I I I I show here the MEP pipeline is to try and tell you you know where weak censorship lives where it comes from so at the very top here we have I guess what I label the private domain right the user and the wallet is kind of stuff happening offchain and so there's not much we can do it's in the private domain and then at the very bottom you have strong censorship which originates from attests right if you have a 51% attack it's the validators the ATT testers that are doing this attack and leading to strong censorship and so the weak censorship is the stuff in between from Searcher to proposer okay so what are some of the problems with weak sensorship and there's at least three of them um and they all have to do with inclusion one is inclusion latency meaning that transactions take longer to come on chain this is something I mentioned it could be a few seconds of delay or a few minutes but there's uh another problem which is inclusion bandwidth uh and and and I'll give more details about that and then finally there's inclusion memetics and roughly speaking um the the less important problem is is actually latency is the one that people think about immediately but it's it's it's the less important one the most important one is the inclusion memetics at least in my opinion now um there's this kind of formula for inclusion latency which is um you take the latency of the L1 which is for example 12 seconds and then you divide it by the amount of blocks which are not being sensored so for example if if only half of the blocks are not being sensored then you divide by 1/2 which is the same as multiplying by two and so your latency kind of doubles um and this is relevant for tornado cash transactions today that are um tornadoes on the o ofac list and very roughly speaking there's 75% um censorship meaning there's there's only 25% of block space which is neutral and so actually you have a multiple of four there uh on on on the uh inclusion latency now inclusion bandwidth has to do with rollups not transactions but more so for rollups if you have a sensored rollup so one example would be Aztec Aztec is trying to build a rollup on ethereum and it's trying to you know have this private kind of mixing service I guess at at um within its rollup and it could you know just like tornado cache get censored now the problem one of the problems with a weak censorship is that if only 25% of block space is neutral and 75% of it is sensored then now Aztec can only have can only consume 25% of the blobs and so its Max Capacity its maximum bandwidth is actually uh only 25% of the of the of compared to another rollup and I think this is an even bigger problem than the inclusion latency right for inclusion latency people will say you know Bitcoin has 10 10 minute block time and so you know a few seconds of delay is not a big deal and you know maybe that's the case but for for rollups and bandwidth it's a huge issue and then there's what I think is the biggest issue with weak censorship is mimetics it's just a very bad look that 75% of etherum blocks are censored today you know it's kind of a sign that governments kind of indirectly have influence over over ethereum and you know when you even just mention the word censorship the first thing that comes to mind is censorship is bad and so it's just not good from a marketing standpoint and it's also been used by journalists as a you know as a a way to to to criticize um ethereum and it's also been used by bitcoiners and various AOC critics of of eum and so I think it is something that we need to fix and the reason we I think we need to fix it is because there's this mimetic chain going on right censorship resistance is at the very beginning of a chain which ends with monetary premium so sensorship resistance gives us credible neutrality which gives us social legitimacy and and means that e for the asset uh can gain the so-called monetary premium which is very important for from a utilitarian perspective why because monetary premium gives us a huge boost in Economic Security as well as a huge boost in economic bandwidth for applications now what is the situation today in terms of censorship unfortunately five of the top six Builders are censoring and uh these are the 7day dominances and so you can see the top two Builders account for roughly 60% of eum blocks and they're both sensoring and so if we look here um just just with these builders that are listed that's about 75% of Ean block space which is censored um now I think it's it's a pretty bad situation because you know we we've had 75% so far but actually the progression has been in in in a worrying direction if you look at Beaver build here um Builder ZX 69 and Gambit just a few weeks ago they were non- censoring so something happened they flipped to becoming uh censoring and so it's possible um that that we're actually going to have much more than 75% censorship in in the in the short term and one of the reasons has to do with Searchers so it turns out that Searchers have a huge amount of power take for example Jared from Subway who is like one of the the big Searchers Searchers provide flow bundle flow to builders and today Jared from Subway provides that bundle flow to both neutral Builders and censoring Builders but imagine that Jared from Subway gets a new job or was acquired and now is Jared from cadel well Jared from Citadel might say Hey you know where Citadel we don't want to send our blocks to non censoring Builders we're only going to send them to censoring Builders well now as a neutral Builder you just have to you know go out of business because you're not going to win a significant portion of blocks if you don't receive the bundle flow from Jared and so there's these few Searchers out there or o OFA providers whereby if they suddenly decide hey we don't want to give our order flow or bundle flow to uh neutral Builders well we're going to see an eradication of neutral Builders and it just takes one of these meaningful Searchers to do so and so a possible situation is that layer one the dominance of layer one neutral block space is basically going to be the non-mf Boost portion the self-built portion which is about uh 6% um so this is where we might be beheading very very soon and it's even worse than that because the portion that sells build is also kind of shrinking over time as more and more validators that are somewhat unsophisticated realize that they're losing a lot of money and and kind of activate MF boost as opposed to self-building okay so I think I hope I convinced you that there's real problems even with weak censorship it's not just strong censorship which is a problem now let's talk about some of the the possible solutions here so there's basically four solutions that I see um encrypted M Pools inclusion lists me burn and andrine PBS and basically there's four different actors that now will have some sort of power over improving the situation so let's start with users users with encrypted uh M Pools will have the option to encrypt their transactions and basically say no no no you can't see the concept of my transactions and that's going to be very helpful for censorship resistance we're going to have proposers who now with inclusion lists are going to be empowered to force include transactions on chain again helping with censorship resistance we're going to have ATT testers with me burn that are going to require that the proposer picks the most valuable block therefore making sure that any block that does censorship um it you know is is much more expensive to get on chain and then finally we're going to have the relays with ENT Trine PBS that will can just piece out and and and disappear okay so um let's go through each of these four um Solutions encrypted M Pools inclusion lists me burn and epbs so let's start with epbs actually because it's very simple the idea of epbs is to remove the relays completely so relay sensorship goes to zero nice and clean um formula here the next one uh which is a bit more interesting is forward inclusion lists so the idea of an inclusion forward inclusion list is that as a proposer um I will self-build a a a block of transactions and this block has to um be some sort of a block template for the next builder in the next slot and this next Builder has the option to reorder trans transactions and insert transactions but cannot delete transactions uh from that list when they're building their block so this is where we are today before uh inclusion list the the the the neutral block space on etherum corresponds to the the block space built by neutral Builders but with inclusion lists we're going to have this additional component the the plus inclusion list percentage so just to give you an example let's say half of the builders are sensoring half of it is neutral and half of the proposers are forcing transactions on chain then basically we're going to have half plus half is one but then there needs to be a corrective Factor when they overlap so 1/2 * 1/2 is one quarter so you subtract one quter and basically we'd be up to 75% of neutral blocks based on ethereum so with inclusion list we can significantly improve um the the layer one uh neutral block space Domin another really really powerful um technology is encrypted mols and there's various flavors of meols and you know there's various talks online that you can find on that but here I really want to look at The Economic Consequences of of meols so today before we have encrypted encrypted meols the cost of censorship is absolutely minuscule it's basically just the sum of the tips of sensored transactions now there's there's very very few sensored transactions out there and they each pay an a very very small tip uh you know on the order of one gay per gas so from in terms of Revenue that's sacrificed from the builders is very very small but what the encrypted moles do is that they basically transform the costs to the sum of all the in the tips from all the encrypted transactions and the reason is that if I want to censor um now I have two options either I include the encrypted transactions but now I take a risk there's a risk that within those you know X um encrypted transactions one of them is is a transaction that I want to censor or I just censor all the encrypted transactions but if like the vast majority of encrypted transactions if the vast majority of transactions are encrypted then now I have to basically censor everything and so the cost of censorship um you know goes up dramatically um you know maybe by a factor of you know a thousand X here now another really cool uh technology is is me burn so um right now this is the same formula as in the previous Slide the cost of censorship is the sum of the tips from sensor transactions but what me burn does is that it adds another component to the cost of sensorship and specifically it adds the base fees of the sensor transactions so now as a builder that wants to sensor I have to cover both the tips and the base fees but it turns out that tips in in in in practice are much much lower than base fees so generally speaking to give you an order of magnitude tips about one gray per gas but base fees is let's say 30 gray per gas so there's a 30X Delta between the two and and if we comound the two technologies me burn with encrypted meols so we have a,x here we have 30X so we basically making the cost of censorship roughly speaking 30,000 times uh larger by by compounding these two technologies and what I believe will happen is that censoring Builders will just go out of business it will be unprofitable to to to to be a censoring Builder once these two technologies are in place and then there's this one uh final idea which I want to mention it hasn't been disseminated very much but if we if we go back to the um to the inclusion list there's a bit of a a problem here which is that the the percentage of proposers that will start using inclusion list might be relatively small it might be 20 30 40% and really we wanted to to be as high as possible ideally 100% and what we can do is we can apply the same ideas as me burn which is to maximize the block but for the inclusion list itself so for example we can have a rule which says that a block is only valid if the inclusion list was built in such a way that it maximizes the amount of Base fees in the inclusion list and so it's now to the incentive of proposers uh to to build inclusion list with all the transactions that they see in the men pool um and basically removing the option to not use the inclusion list and so I think we're in a very very good position from a from a research perspective and from a design perspective I believe we have all the tools uh to fight weak censorship it's just a matter of time until it it's it's solved and that's it thank [Applause] you