Sebastian Bürgel 🛸 The Dark Endgame of CL, EL & Application Layers Without IP-level Privacy

[Music] so first of all I would like some stats here uh who has been using any evm chain in the last 30 days okay that was a mandatory arm stretching exercise now um how many of those people have been using exclusively their own notes when interfacing those evm chains okay no RPC providers in the room good now as you do that right so as you rely on other people's computers to interface any evm chain you know who do you like as it turns out all of you actually trust and there's a narrative which is not new is it maybe this company um I want to be very clear that this is not meant disrespectfully I'm very grateful that there's somebody who runs infrastructure that we all are too lazy to run ourselves but you know maybe you don't trust that company in fact actually trust that company or that guy or that guy well you know what does trust even mean right so trust in this context of like we trust in fura it's all evil like what does that even really mean so firstly it means trust not to censor you and um how do we you know how do we get that trust well luckily it's not that bad right luckily it's not that bad because all of ethereum most of it anyway uh relies on free and open source software so you know I can inspect that code um modify that code and run that code myself that's kind of cool but actually uh it also require standardization so I'm kind of grateful that we have this whole EIP process and people standardizing things because it means switching costs are low and thirdly and we heard about it a few times already retail Hardware requirements are pretty important because it's not just you know one of the very experienced gentlemen that I showed before but every one of us on their dub note at home who can run a note that's cool now secondly um is trust to deliver correct data and uh that of course makes sense and the obvious solution is again you know run your note from home but as we've seen that all of us are too too lazy to do that and you know if we didn't change that in the first eight years maybe we're not going to change it in the next eight years um there luckily is a solution and that is light client verification again mentioned uh but I would say here in this room can we please start using that uh also a shout out uh vitalic mentioned before Helios from a6z crypto I know it's great that these big VCS push that forward but there's another one called kefer Dosh um yeah run by shest and a bunch of his friends uh which is very cool and uh yeah very efficient now finally and that's uh the part of my talk that I would like to dive deeper in is trust to not track your every move that you take on a web 3 and as it turns out that is pretty broken today and I'll go through you know what that means in a more broken down fashion but basically strong fullstack privacy is a necessity because how much does it help you that nobody can you know technically censor you on accessing the the nodes nobody you know can deliver you fake data when they um Step one step behind you and observe over your shoulder uh everything that you do um yeah so specifically let's look at let's look at it from you know kind of a a layer stack approach right so let's start with the application layer um so we'll look briefly at tracking another web to ugliness later on so I think we have to subscribe to some values and you know agree that user tracking is evil because if we don't agree on that a lot of the other stuff you know it it just all breaks down it doesn't help you if we have perfect consensus and execution layer privacy if then on the application layer we track the hell out of users um but that being said there's of course problems on the execution layer right so onchain privacy is needed again we have a bunch of these things going on um yeah but again uh let's please use them uh and then interestingly we have P2P level leaks and uh this is not just on the peer-to-peer layer of the execution layer itself but also kind of the auxiliary Tech around it right solvers Builders indexers um that's all infrastructure that is uh leaking uh information that can be tracked of individual users and finally of course consensus layer um yeah bators are not private and unfortunately I know there's great movements made to a single secret leader election but that does not really fix privacy concerns cool so luckily um we're not alone right we have the old world which we try to decouple from that protects our privacy and I think it's fair at least for half a second to acknowledge that so we have things in US Europeans that have something called gdpr there is you know those Californians that have CCPA but it's not just State actors it's also um corporations which um yeah kind of self subscribe to privacy policies sometimes better sometimes worse um yeah lots of proof for improvement there but you know as this old world admittedly like I brought up these names of this organization and the people running them right and my point is you can enforce privacy um you know privacy regulations because there are identifiable individuals incorporate players which act in a space now what do we do we decentralize all the things right now what does that mean right that means means we remove these trust guarantees that are provided by laws and regulations of countries states and private for-profit corporations we're replacing them by trusting all runos on the internet to not data Harvest everything they can about me to not use it against me in the worst possible way that is not great yeah so that is not great and my point being here is Yolo decentralization without regards for privacy is a problem and in fact it makes the trust assumption with regards to privacy much worse than the centralized world that all of you get away with today and um yeah I would like to dissect that one a little bit more to drive that point home and then think about what we can do to fix that so connecting to a non-private execution layer uh again allows some randles on the internet which run that infrastructure to link all your ethereum accounts right people here have a poop well cool because there's infrastructure providers maybe at some point very decentralized ones to which you have no trust assumptions whatsoever which can link that you know pup account your enss account to your savings account that's broken there's infrastructure providers which will be able to link that to offchain identities so I find it actually totally fair that gdpr in Europe considers the IP address personalized information right because you can totally link that to an email if you have a single form where you subscribe for example with any of the providers that you all use use and finally and that's something you know I know there's a bunch of me folks in the room um that I find very interesting these providers see everything that you see in a truly decentralized world without before you see it right that's pretty interesting now it all boils down to me um again I work on the I work on a transport side I work on a P2P side your IP address even if we were to go forward and think about perfect onchain privacy you know your IP address is going to be your primary identifier on the web 3 for all of this above which is something that yeah I'm happy that you know Juan and vitalic acknowledged that before but I think we have to talk more about the specifics and solution space here cool so how exactly um can you then screw users on a web 3 that is kind of YOLO decentralized and uh having not many regards for privacy I'll walk you through an example a fully decentralized app. Unis swap. eth okay so I guess most people here interacted with eth domains uh what's the Tex de for this hypothetical example that again I hope for right I hope we do get to a fully decentralized world I just think there's some problems so how does the TX de for that look like so it's enss obviously right that resolves to ipfs and once you load the page there's the graph delivering data and then you use decentralized RPC provider such as for example pocket um yeah to serve requests now what happens first um our malicious actor is going to um run a bunch of ipfs nodes in fact you don't even need to run the full ipfs nodes you just need to get exposure to the DHT and you are identifying active users by harvesting their queries you're saying okay Sebastian congrats you learned that I'm accessing app. Unis swap. e what do you get from that well I prepare for my next step I know that you are coming online and I want to Target you so um I am ing up a bunch of graph nodes um not and that's what you know a lot of people give attention not to feed you malicious information that's not the goal here the goal is just to De anonymize you and harvest data about you and this is pretty bad right because again on this Unis swap interface you click that drop down like vitalic showed you and you know people correctly identified as Unis swap uh you you click on these drop downs all of these drop downs load data right so there's somebody feeding you that data in a properly decentralized world who knows every button you click even without tracking like before you see it right you don't need to be a super powerful AI um you know to figure out what's going to happen and finally of course you can front run the men poool one thing that I find fun uh and just wanted to mention it's not just the transactions that you can Harvest if you've access to an RPC provider but also uh benign looking things like eth estimate gas right so if I know that um you know some guy might be very reliably sending transactions I'm not I don't have to wait for the transaction to pop up in a man pool I can just look at the E estimate gas because I know you know the lady is going to click on the confirm button anyway before you know she sends it out so yeah that is something which uh you can totally use to front run the mle in a truly yolow decentralized uh web 3 finally again this is not meant in any disrespectful fashion I know that all the mentioned projects are aware of these issues um and are working towards Solutions but this is an example for what we need to consider uh as we decentralize all the things so what about consensus layer privacy uh just a brief summary of an experiment that we run um we've been running a modified version of um of a lighthouse node again uh our code is freely available here that did nothing else but uh collect all attestation data and linking pup keys to um linking pup keys to the um IP addresses and we've managed to correctly identify after a couple of days uh all uh Hopper team members which run uh validator notes on the nosis beacon chain so yeah I mean uh that is a problem and that is obviously a problem because you can take out a note if you have correctly identified them so I get this a lot right Sebastian is this really going to happen so a is privacy really a problem and is anyone ever going to exploit that so let me just quickly answer both these sub questions firstly um this this is today's said State uh of ethereum Deb privacy and to be clear yes I do want to call out Unis swap here um yes I do want you to look at the network tab when you interface deps right small D so this is happening when you interface Unis Swap and the interesting thing is this is not on D on on app.

Unis swap. org this is the ipfs deployment okay in the ipfs deployment they're uh logging stuff such as my wallet address they're linking it to a un unique thing called device ID well where would that be used oh it's used to link all my different requests that are being made you see what I was attempting to swap there when exactly you see every single freaking button click in there like you even know the exact version of metamask that I'm using so if there's a zero day in metamask well congrats there's a bunch of devops guys which have full access to you know all my vulnerabilities please let's not be like Unis swap so the second one is and Tony right after me is going to go much more in detail of the very sophisticated analysis that's uh happening there in the Mev landscape but to me what is going on there this is a data harvesting MVP on ethereum right we see that this is basically a billion dollar game and it's happening so the question is is you know to me this is just data harvesting menol data this is again the tip of the iceberg which is a very small subset of the data that you could have access to in a fully decentralized fashion and it's already a billion dollar game now you know um there's something we have to do about that to not let that game just grow in an in an entirely unbounded fashion so and you might ask the last question well okay but are people really going to exploit that we know who our Most Wanted next ethereum users are we want the institutions right don't we want the institutions well look at what the institutions do and think about the cute me poool that we are harvesting today isn't that very cute compared to hedge funds that are literally harvesting data of millions of people like harvesting dat data off a freaking space yeah so you think that transport layer information is a tinfo head future no I don't think so I think it's inevitable we will harvest we will have actors on ethereum that will harvest all data that is available and decentralizing things makes that way easier for those actors cool so again uh that was kind of my conclusion uh strong privacy tag is our primary insurance provider to prevent exploitation of every single web three user and actually we're kind of aligned there it doesn't matter if you're talking about the individual or if you you know want to protect these trillion dollar hedge funds from one another we are aligned that we need privacy to make ethereum a safe space to interact and just to wrap up I I initially showed kind of um some of the issues we had on the application layer like tracking right so we have to defend the core values um I know ethereum are nice people and that's something I personally really value and appreciate but still you know we have to call out the actors which are not aligned with our values uh so I take that I take that start here today uh execution layer onchain privacy right that's cool but use them integrate them and more importantly we need support on the legal front I know most of us are tech people we don't like to talk to lawyers but I know at least two projects on that list that came to me and said hey Sebastian what should we do my team is afraid to work on privacy this is an issue right we need to we need to provide guarantees for the developers of our magic technology that they are safe to build what they're doing finally P2P leaks um yeah solvers Builders indexes and so on mix Nets are great um I'm not going to shill what we're working on I'm just going to say Hopper is a mixnet we use that and uh we specifically use that as middleware uh on the RPC layer I think RPC um middleware is an under appreciated uh building block so I just also wanted to give a shout out uh we heard I think also of italic say before you know uh how do we deal with forced inclusion well I would say forced inclusion is again something that shouldn't happen through a user front end this is something that should happen through an RPC wrapper and I hope somebody works on that because I want to use something that's truly censorship resistant and finally consensus layer yes validators are not private and ssli doesn't fix that but strong privacy teag does on the flip side I know latency issues are hard so that's a reason why we focused first on the uh RPC middleware stack working on RPC over Hopper and if anybody wants to support that or use that please come to me afterwards and with that thank you so much and thank you for jointly upholding the values of ethereum in a bare Market that's easier please uphold the values when a bull market is coming around thank you so much