Protecting Web3 from Bots & Bad Actors with Open Data with Disruption Joe

doing seems to be working yeah cool uh behind that is a few million dollars spent in fraud defense uh services and we're going to take you through the Journey of how we are going to let you protect your communities for things that are different than one token one vote right you want to do one person one vote quadratic funding quadratic voting you need civil defense so with Bitcoin we're just helping you fund what matters that's our big thing right here uh we create solutions to empower your community to do that and we're going to talk about what you actually need because there's tons of tools that let you allocate Capital right we have plenty of Dow tools let people make some decisions together but you [Music] thank you what are the rules what is the Smart contracts set up for how we distribute the funds well there's also people in Dallas is anybody here in a dow and you have to distribute the funds in Fair ways okay that's more allocation of funds you could think of the allocation of funds using one of these mechanisms within a corporate Innovation Department you think of a government using this mechanism if you participate in making or designing the rules that more than one person come together to allocate a pool of funds you're a program manager in our marketing terms so there's that it was well thought out okay it was a program manager do they fund what matters your community has a pool of funds it needs to be democratically distributed so that means it's not just Evan going and saying this is how we're going to distribute it and hopefully it's not just Evan and Pete going hey this is how we are going to distribute it but it's going to be some form of all the people in that Community who have the legitimate right to be there and have a decision in that because that's what consent of the government is what kind of fairness are they looking for well they're looking for consent of the government they want to have a voice in the Democratic process they want those decisions to be community and when a decision gets made or rule is said to be broken and they say we're not going to count your vote well they want that to be applied equally across everyone in the community right this is a new opportunity to redefine those systems where just because we don't look the same or we don't have the background that's the same the law gets applied to you differently so when we talk about gaining a community's trust we're talking about Bitcoin Building Systems that allow you to equally apply the law across behaviors and ban behaviors not people a few other things that you're going to think about are like who's in this community right we have this term we call sybils and if you're in web 3 have you heard the term sybils yes okay keep it up if you know what that means okay so we want to say contributors are valuable to the ecosystem or they're not but it's more complex than that you have type 1 and type 2 errors that you're going to run into and this is the problem that many communities are facing they're trying to figure out how to identify sybils non-valuable people versus valuable people but what's actually happening is we approach the type 1 type 2 errors because of the way people interact with this for example our first personality is a noob how many of you are noobs now Brave love it uh new to web3 how much unchain history do you have as a noob right maybe when you found out about it you had one friend who said hey come do this donate to this this is why you want to be a get coin if if the ux is terrible you probably aren't going to fight over it right and you wouldn't pay twenty dollars per transaction right who here has paid twenty dollars per transaction man like yeah I guess whatever yeah noob's not gonna do that okay uh so we start to identify what a noob looks like right and they're valuable to the ecosystem we could stop every single civil attack today by cutting off access to the noobs right they're the ones that look like fake new accounts so we have this quadratic voting mechanism quadratic funding mechanism that gives more preference to the many than the few and wealthy but to do that and get rid of all of them if you think of it just as stopping bad guys you're gonna disempower the people the algorithm is literally created to empower so it's very easy to do it wrong when you're trying to do something like civil defense because what we're doing isn't defending what we're doing is actually empowering communities to have Fair elections and fair votes in how they determine how they're allocating capital and this all happened behind the scenes in a proprietary way with Bitcoin in the past we had to go out and distinguish these personalities a farmer they're not malicious but if they donate a lot to get coin grants hoping to get an airdrop well they're going to skew the way the vote allocates you know they only care about when airdrop but who who the [ __ ] am I to say like they can't want an airdrop like somebody in a place in the world where they're not earning nearly as much as we are here that could be life-changing money for them and I'm just going to go and say we're going to bend that person from the ecosystem like [ __ ] that that's not what we're doing the next thing we gotta look at is well who is malicious attackers are malicious right they're actually trying to steal your funds they want their donations or their votes to count and they will jump through the Hoops to do it Hoops like collecting our passport stamps or whatever nfts you say they have to hold to be counted they're likely to only fund their project they have maybe some grants that they're trying to allocate funds to they may actually study for weaknesses I've seen plenty of testing that they've done just running different algorithms trying to hit the network directly going around the UI uh every every round they test new things we develop new answers and this is the fundamental problem right what we have today is one token one vote it's not exciting we know it's not the only potential so how do we move to this new era where we can have high resolution democracy that prefers the individuals over the wealthy right many over the few here first thing you gotta say is will they attack yes they will attack okay we we call it the fraud tax this is like if we didn't take any measures and then after we do take measures what's the Delta and you see that a few grants lose some money but many grants gain a little bit back that they wouldn't have had if people were stealing those funds we got to Define who we're going after too the parts of the Bad actors here this is uh round 14 and I think every Edge indicates a thousand transactions so there's a lot of data here right it just crashes your browser every time so that's why I went with around 14 it's around 15. but we got to get through that and find some of this bad behaviors this is people who only have one difference in each letter it's a thing Omni over here brought in is as a data scientist he brought in the 11th steam distance saying like if I make account disruption Joe one disruptor Joe two disruptor Joe three disruptor Joe four and we all donated to the same grants the same way it's probably simple right however disruption Joe 69 is a really funny guy and he had that for a long time right three years he's a native we want to make sure we don't classify him wrong either so we have to identify a bunch of single behaviors this uh was made using Rafferty Open Source Time series graph DB really cool tool group called promptory words with us on it you're identifying behaviors motifs that are anomalous or seem Sybil and then you say oh maybe this wallet is something that should be flagged based on this one Behavior we need data scientists to come in and say what are the anomalies uh what you're looking at here is some representations of our donor DNA where during the round we look at every single Grant you could have donated to and we look at the anomalies in that or how close you are to other donors in your funding pattern here's an idea where we got to say okay how can we compound these like on the grant side we've done a thing called on-chain intersectionality where we look at very high quality signals like if you bought a thousand dollar premium subscription for Bank list you have that nft probably not Civil Right you spent a lot of money to be a symbol if you did now we look at the wallets that we call burners that have nothing they have no nfts no transactions they have clear preferences now we can go back and look at the donors and say if you preference the grants that were preference so much by the ones that we know are probably illegitimate that's what's happening and then we take each of those behaviors and we put it against our Thor and our Loki sets these are our for sure known sybils uh for sure not civil so we try to say like all right is this specific behavior actually giving us something and doing this we've stopped three million dollars in fraudulent funding since uh grants round nine in December of 2021 as a Dao uh we also at get coin just launched a protocol our aloe protocol we launched passport we launched a grant stack which is four different interfaces for individual users in their Persona to interact with those protocols so building this as a dow has been one of the most incredible things but we started proprietary back in the day we didn't really do civil defense at all the first rounds we were just kind of hoping we started building reactive tools eventually ZK sync made the get transactions so cheap we needed to figure out how to scale that hired block science I call it the BS era before civil defense like we we literally did nothing and then started realizing there were problems and this is what every Community or program manager will go through if they don't use that great ability we have as humans to learn from what we already did the next era we started using machine learning so we could scale like human evaluations with human in the loop semi-supervised reinforcement learning we launched a trust bonus to weight users as they came in based on having other identity attestations we tried we launched the Dow and made it so the community owned the subjective decisions that are made along in this process and then we got a little curveball when they introduced these pause rounds which are like climate change because all these noobs came in who didn't have githubs which was our primary identifier so now all our data is kind of like not what it was before it doesn't tell the same story so this was like the proprietary Defense era where the grants were reviewed for eligibility we hired black science we got the humans in the loop going but it led us to the most recent things before the protocol the last era is where we know the protocol is coming the product you might have used it and it was like oh man this is clunky it's because we were really like let's throw that away let's get to this aloe protocol here that means we were doing on-chain analysis we we decided we'd launched the open data Community where we could actually crowdsource more of these things because the protocol is gonna be open so sea grants was like this crazy Legacy code thing and people didn't realize how many different levers and pieces were in there for you to make a decision so the one thing I want you to take away today is to grow your community you need legitimacy and the reason people look at get going and say hey that's a cool tool we want to be able to allocate our funds for our community like that is because they know they can trust the outcome that was a lot of work and that has all been made open to you kirkhov's principal's idea in cryptography that even if the attacker knows exactly how your systems work it still stays up it's more expensive to attack it than to defend it and we're going with that model with open data we now have our passport which gives you a transparent and auditable history to how you applied algorithms to somebody so you can use algorithmic policy in a transparent and audible way it lets you gate or wait access to any decision making we have the Civil scoring Legos that says we were running machine learning but we lost comprehensibility humans didn't know exactly why but we can have the benefits of machine learning for the accuracy but then each Lego that's open source Docker packaged and put out there so you can switch into the your machine learning algorithm very quickly this lets a non-technical round manager say oh your account wasn't allowed because you received your initial gas from a wallet that also funded initial gas for 10 or more other wallets that donated in this round you had over 200 transactions in under 30 minutes at some point in the history of the wallet is obviously uh is obviously that this has done some bad behavior and now we have this open data Community even innovating dashboards so you can plug and play new LEGOs because civil resistance is an infant game that means it goes on it's an iterative process so when we do something new to stop them coming in the front door they try the back door right so where Kevin McAllister here having to come up with new ways to stop them constantly and we think we can do that even with them knowing how we're stopping them by crowdsourcing and bringing you all into it by having open data on the protocol and the economies of scale benefits between you and every other community this open data Community can continue innovating for the new solutions that are needed and the point is if you look at git coin you're like I want one of those I want to be able to allocate funds in the way that they do with a trusted outcome now that we've released the protocol you get economies of scale benefits you get the benefit of every other person in here going through the same issues and you get the benefits of knowledge that stopped three million dollars in fraudulent attacks over the last year and a half and that's in your toolbox to give your community trust that you as a program manager are making sure that the outcomes are legitimate and they don't have to trust you all our processes and tools are fully open transparent auditable and reproducible find me another protocol that does that and I will gladly hands down walk away but that's why you want to allocate your funds with the aloe protocol that's it foreign [Applause] [Music]