Demystifying DeFi Risks with Statistics and ML / Marijo Radman

Transcript

you can hear me nice okay great I believe most of you here in Rome can agree with me that decentralized Finance is still a very dangerous place to be in even if we just look at the recent events starting from the last summer when Terra Luna happened followed by the FTX crash we had attend the usdc Deepak happening in combination with the Silicon Valley Bank and lastly also the Euler hack which reminded us basically how dangerous the space is and we should ask ourselves why is this happening I mean we have those companies who are doing the

security assessment like Auditors on one side we have on the other side companies who are focusing on the economical side mostly on the protocol development side and yet these events are repeating constantly and I believe a more important question is actually when we look all these projects in D5 they are like mostly two to five years old so basically unproven projects managed to attract almost billions in capital and we should ask ourselves what is the criteria for those people who deploy their Capital to these protocols and when we look now D5

I believe the idea was to build our open transparent more secure Financial system alternative to the traditional Finance piece and yet what we have today it's a worse version of that because those kind of events are not happening in trade 5. and if D5 should go mainstream we have to rethink how the current space is working especially around the risk ladies and Gentlemen let's introduce myself I'm the co-founder and CTO of solitary Network I spent most of my career in Corporal development working for major German corporations like auditative or aeon I'm in crypto since 2016 mostly working on the device side network was co-founded beginning of the

last year we closed our first financing round in September of last year and since then managed to build a team of 18 people out of which 14 are working on the development side we are headquartered in Munich and have also offices in London and Zagreb so let's try to understand why do we have these kind of events and let's just take a very very simple example here I believe most of you know these strategies which were very popular in the last Bull Run so basically you have your stable coins and you put it on some of those protocols working to get additional yield in this example we have two lending protocols on one side of any

other side compound and what is actually the mainstream doing what was the mainstream doing how was the allocation happening uh to those protocols most of them were just looking at these liquidity pools looking at the yields checking where is the highest opportunity and in accordance to that move their Capital there in this case I have a seven percent compound is nine percent so basically they move their stable coins to compound I guess most of you hopefully can agree with me this is not the smartest way to deploy your capital why because we are not taking the risks in account

what are the risks behind other what are the risks behind of uh compound a better way of capital allocation would be actually to look the risk reward ratio so basically taking the risks taking the reward calculate the ratio and then in accordance to that deploy your capital and now we are coming actually to the problem of D5 when we look at those protocols here we have actually also the chains because RV is for example on ethereum of course a compound also but it can be also another chains for example Avalanche and you have to take this also into account

between the chain like ethereum and the protocol there is usually a L2 solution because of scalability issues like for example arbitrom and that basically describes the inner working of D5 basically D5 is nothing else than a composite of various Primitives and actually the thing is much more complicated because this is a very simplified version of how D5 is working we actually have over 100 protocols and they are all interconnected in this case we can also take the bridges in account especially if there's some kind of cross chain strategies in place we have to take the Oracles in account because for

the correct working of these Landing protocols real data is also needed and oracles are providing that and then on top of that we have then yield Navigators yield optimizers who are moving the capital in accordance to a predefined strategy now the thing is like all these Primitives here I mentioned have their underlying Dynamics they are underlying risks and basically this is looking like this so basically when we start just with the chain we have governance structures how much distance centralization is there we have validators slashing events happening we have

consensus mechanisms who are vary from chain to chain then on top of that L2 comes with their governance structure with the sequencer then the D5 protocol besides the smart contact issues we all know we have to take care also about possible liquidation events happening there there are liquidity pools you get inside you are actually not sure you are able to get outside of it and then on top of that you have all these yield Optimizer seal allegators who are bringing additional security issues and additional centralization if then bridges are involved then you have finality of one chain the finality of the second chain and the bridge basically making the whole situation

basically picking the worst of this tool um of course and lastly you have oracles with their data feed inconsistencies so basically a lot of those attacks happened through the Oracle side and the recent discussions we have on the uh Innovation side on the consensus mechanism like shared security risk taking and so on is going to magnify even more to all these Oracle side and so basically What's Happening Here is um if one of those protocols breaks like for example compound it will affect immediately the other Primitives like immediately we will see some kind of consequences so novel like liquidation events some other Protocols

are going to be affected and the question is like how is possibly a user able to get grasp what kind of risks are he is he taking when he deploys his capitals when we look the current space and when we look these risks so basically those risks I described you are nothing else than technical fundamental risks which are unknown in the traditional Finance space so basically there is no real Frameworks standardization how to deal with them current Solution on the market who are trying to grasp that are usually very simple static and subjective so a very simple case is basically how they assess smart contract risks they are

just looking is the Smart concrete audited for how long is that contract deployed was it ever hacked and that's it on the other side all this stuff here is happening like dynamically like we cannot just take the smart quadrant in account all these liquidation events changing like from one event happening to the other one and that's basically where solid networks come in we are using a big data approach so basically by quantifying the risk in real time through indexing and monitoring monitoring the of the whole ecosystem starting with L1 L2 chain with the D5 protocols stable coins Bridges oracles including the liquidity pools

we are focusing on the fundamental side of the risk so basically Gathering data around security around liquidity flows the centralization governance socials and development activity and enabling our customers to build risk related products out of the out of it and the most important features of course that those outputs should be like in real time we are also aware if those Solutions should be also used on the device space we have to decentralize the text stack because we want to help these D5 protocols to be like more advanced and not to be attack Vector on them as it could be the case so basically how should this be solved this is a very

high level uh overview of our architecture just don't be like scared about that it's actually very simple so basically in the first step what we are doing is like we are running our own nodes by gathering the online data and through our data Pipeline and running Lambda functions to get the opt-in data and what we learned from discussions with many protocols to enable decentralization of the tech stack we head to red thing the whole idea of what does it mean since the centralization and the reason for that is basically most of these projects are in a different stage of decentralization and it's even more interesting most of of them have a different understanding of the centralization

and it's getting even more funnier when you start to talk about the cost benefits when they realize that the centralization has to be paid then we understood okay let's provide a solution for all of them to get their own way of root how they want to decentralize what does it mean so basically they can decide do they want to have a centralized way of getting the data or decentralized do they want to are they fine with centralized immutable databases or do they want to have decentralized Storage Solutions integrated in their data flow how should the computation done is it often computation enough or do they need some kind of decentralized network on

that side and basically the whole route is given to them on top of that we are going to enabling like running various algorithms from statistical models network analysis up to machine and deep learning models depending on the use case they want to build and what we learned from our customer discussions what these Solutions are usually used for are for example customizable risk costs which update itself in real time they are used for portfolio rebalancing and wall strategy or the basements this could be like Risk Frameworks for insurance protocols or liquid staking or just a monitoring or alerting system in place if there is for example a new

ecosystem chain involving or on a thread file uh trading infrastructure in place and basically the output is of course through apis if the protocol of the customer is comfortable with that but it can be also through oracles if he is more under the centralized way around and here are these customers I mentioned before so basically on the D5 side these are mostly like ancient asset managers like hilligators Shield optimizers who are living dealing with liquidity provision Insurance protocols decentralized exchanges but also protocols we don't know at the moment and actually the idea is to open up the platform that was actually just the

first part what we built but the idea is like also to incentivize data providers to provide unique data which we don't have which are necessary to build those Advanced risk Frameworks we also want to incentivize these routes so basically as more customers are using a very specific route as cheap as it's going to be in the end we are also going to incentivize three third-party providers of models what does it mean for example if we want to have also agent-based simulation models there are companies who can integrate their processing infrastructure into our solution to build like even more advanced Solutions out of that basically getting the outputs from here combining

with these models to get something totally unique on the risk side and lastly we are going to incentivize everyone to build risk related Frameworks and basically also with the idea to capture use cases we are not aware at the moment uh and all of this is going to be like paid to the customer through a consumption based model from experience of other protocols in the space we understood from the beginning it's not really a smart idea start from day one it's decentralizing the tech stack and opening up the platform as always said okay let's build a centralized solution the first part

provide immediate value and capture this customer who are fine with the centralized case and I'm happy to announce that we managed to build our beta solution which is now live with our first paying customer and it's now going to improve like even more with our 30 design Partners how does it work uh we made it like from the day one as user experience most important thing uh it should be like as easy as possible so basically if there is no knowledge about risk that they can build a very simple risk framework in this case block split risk framework from conference a week ago the idea is like basically he just selects The Primitives he's working

on and out of that a default risk model should be deployed without thinking what's happening so basically these default risk models are trained by ourselves and our approach is basic to build various risk models using various methodologies on security side liquidity side with subscores and this overall score can be then combined out of those depending on the formal user is fine in this case it was the average then if the user wants to optimize a little bit without going too deep in a launch you can change the weighting factors of this sub models to get even more unique case for his use case and lastly if the End customer has

domain knowledge expertise in D5 space he can go on the parameter list he can change the weighting factor of every single input parameter for those models to get something more unique on his side of course these models watch really deployed are going to be open so basically he can then optimize on that side and lastly also the models are going to be adjustable in the second piece as I mentioned we are now live from here on we are now going a direction opening up the platform that third-party providers can provide the data build Advanced risk Frameworks but also decentralizing Tech stack with the idea to launch our product in mid of November Ave at devconnect in Istanbul

um so when we look the market space at the moment I mean there are a lot of these risk related companies and so basically this is our definition we know there are many companies who are focusing on the protocol design like companies focusing on the security side of the protocol companies focus in the economics of a protocol yet the liquidity provision the capital provision is not really addressed in in a right way we know companies on the economical side these are usually companies who have a team from traditional finance and they are trying to build these models into D5 space and that's basically this field where we are is in our opinion underserved is basically this newly introduced risk

category like at the fundamental side of the risk where solidity is focusing on as I mentioned are we targeting the D5 space liquidity position and the traditional Finance space mostly hedge fund managers we launch our product with the 30 design partners and if any of you here in the room is willing to test our solution just ping me after the talk and just to conclude I believe the tweet from Dan Market still in time this was from the last year so basically the D5 in current form is going to change and we believe risk management is going to be like major Drive of this change and solid network is going to enable that

thank you cool so do we have any questions for Mario from the audience hi um Carol we actually developing kind of competitive solution but we started with the end user from extension also the training model I'm curious so how do you plan to integrate with the defy projects so uh because you you mentioned it's going to be like Oracle and API and so what's going to be the main focus I believe this is like really depending on the customer like you did probably also your customer Discovery and you've

learned that some D5 protocol Protocols are very sensitive how to get the data and how deep they want to go inside of that some of them are comfortable with apis if they just want to have this information in the front end it's a really BD topic and we build this infrastructure routing layer with the idea that we can provide this decentralization depthness depending of the use case they are following and in the end I believe the most interesting part is going to be once they realize the cost of other centralization what route they are going to pick up so I have a question um what will decentralization look like for solity

the centralization facility I mean we are going to build this overarching layer to enable that all these Primitives come together on the middleware side so basically we are not going to build Storage Solutions we are not going to build decentralized Computing networks we are not going to build like triggering Solutions of all these Primitives and oracles but we are going to bring them together and we are going to focus more mostly on the incentivization side enabling that all these Primitives working together in a right way and that means of course we are going to create that entity opening up the platform create create governance structures but this is like a process which has to be happen over time so that's the reason why we are very

careful around that and so what is your risk assessment for Mika and other regulation that we're starting to see spark up in the space what sort of time frame do you see that you have to to build out before potential regulation is is an issue for solitary I we are aware of that so basically we started with all the spontational thinking from the day one about those kind of topics um I believe the most important topic is going to be like the utility what you're building behind this structure and we have this centralized case so basically we can create onboard Customs create revenues on that side and once this incentivization part comes

from the legal side is solved but we have to see how this all is going to involve so basically it's going to be like the company in this form as it now it's going to change of course um it's going to be a network and we are already building on that side great thank you so much for coming thank you just the last question if you have questions there was a QR code feel free to join our telegram group and ask questions there so we have five minutes uh until our next speaker foreign [Music]