Removing the Human Factor from DeFi Risk Management

Transcript

okay thank you good afternoon everyone so yeah I'm your own Veller I'm here to speak about defy risk management so defy risk management is very broad Topic in this talk and this is also most of our activity it's about the economic risk of things where most of the you know risk activities we see are Focus mostly around you know smart contact bugs and hacks however as defy defy products and Protocols are becoming more and more complex there are also a lot of economic uh risk factors that come into place in this talk I will mostly focus on

lending markets we're in landing markets the main risk factor to determine is the so-called loan to value ratio which basically means how much for example uh US dollar you can borrow against uh for example a Bitcoin collateral or is collateral and this is a key metric in defy lending markets because it affects both the security and solvency of the protocol but also has great impact on the user experience uh some background on myself so I'm the founder of B protocol which is a defy protocol that is live for almost three years and aim to mitigate risks in

landing Market I'm also a lead contributor to risk Dao which provide risk analysis services to different platforms previously I was CTO of kaido network a singaporean-based decentralized exchanges we built the erupt Bitcoin my history with blockchain started already in 2016 after completing my PhD in computer science I was led a building a lot along with other co-founders the first ethereum decentralized mining pool back when ethereum was still a proof of boxing my background is technical I whole PhD in computer science and has a lot of technical development experience

so yeah when we want to mitigate the risk factor so it's kind of appropriate to say what it is we are working on on the last Euro to so recently especially in the last year we're giving a lot of services you know risk mitigation services to a bunch of defy protocol so so this kind of aspiration to mitigate the human factor is also kind of to replace ourselves with some sort of uh more decentralized but even more you know like a smart contact a way of things and risk management as you know kind of notorious history

beginning with traditional Finance you know some hold it responsible to the 2008 credit crisis it is very biased or Park political uh you know in traditional risk management it's mostly about rating the depth and the quality of the depth of companies or Sovereign Nations it's notoriously you know biased toward us D for example and again it was one of the kind of main reason actually to why it Bitcoin right came to life as a kind of an anti-movement to the 2008 credit crisis

but then okay crypto came but you know risk management is still managed also in crypto by humans uh and we also you know what happened with uh cereal Capital going insolvent uh because some of the creditors decided to post liquidation because they thought you know it might be the best kind of commercial or even trading decision to do but it resulted you know like in complete bankruptcy FTX for opposing you know liquidations to Alameda which was you know of course like a total fraud not only build risk management but but you know like eventually when you have humans you know having kind of this power to manage these things this

is where you end up with so yeah so it's not only in traditional Finance we say it a lot in crypto but then of course you know in D5 this would never happen right because Defy is all about transparency so no so if I land in markets are very transparent they are very decentralized but it doesn't mean they are not political and not sort of biased toward human decision and one example here is that you know so many of these Landing markets have kind of their own third-party risk man risk vendor that kind of recommend you know risk management activities and we saw you know this is only like uh five months ago

the risk vendor recommended to freeze liquidations because of certain knocking conditions knowing it could result in some bad debt you know also in other markets and and you know like here kind of the complaint is neither for the landing Market nor for the risk vendor which you know like very professional and smart entity but it is you know one way sort of close there to you know freezing liquidation this is what you know how FTX end up with uh a more recent case uh I don't know how many guys here are kind of following what is actually happening you know in a defy that was like very hot topic uh

well care founder actually took a huge depth position in Ave against its own native token and near gunland which are the risk vendor you know head like one position of you know what should be done but it's kind of representative of the avedao community you know had a different position you know but both positions there's no like actual right and wrong here right both positions are super professional a super kind of honest you know and then also in the middle you know you had like the actual user who was using a D5 platform and you know now there was a proposal to kind of change the rules uh

in order to you know kind of handle his specific situation so again there's not necessarily right and wrong here it's just not what you expect you know from a defy protocol it's not like how I know uni swap you know would behave more generally and you know kind of coming from the trenches and managing you know risk for over two years in D5 protocols so when you decide on you know like on this loan to value ratio you know you have in mind a lot of quantitative and measurable metrics but you also have a lot of you know kind of domain specific knowledge for example assets like usdt you know

like kind of objectively very stable over the past years but you know it is sort of valid to have maybe some I don't know biased opinion against them because of their structure but it becomes you know like even worse when you start to potentially also analyze uh you know like the users you operate with for example it's not maybe uncommon to analyze like the user activity and then maybe you can decide that okay if the user operated mostly in Asia working hours he is based maybe in a certain area based on that maybe we want to give him you know like a worst uh wall so maybe better you know risko so

so all these things you know are kind of very common you know like in traditional Finance but it's not something you would expect from uh you know non-biased an objective D5 protocol so yeah so in in B protocol and also there is our vision and you know how can it be done differently so we aim to replace all these risk vendors including ourselves with a smart contract and now this smart contract will get a quantitative data such as Dex liquidity asset volatility Etc as input we're actually even bringing this data on chain is not

trivial and this is a big part of what we do and then to take into account because you know like division is that eventually everything is fully automated fully objective no human input but you know like to be realistic in the short term we have to you know give some room into some human input uh so here like every entity could specify some confidence Factor so we if you have less confidence in usdt then usdc okay you are also afraid to kind of specify that and more broadly maybe a dog can specify it so we do give some room you know to human input but this is like a single number input which makes everything super transparent you

know users can compare and understand you know the different risks and confidence factors the different protocols take now given this input this manual tract will just output you know the recommended a loan to value ratio also other parameters potentially interest rate Supply and collateral Etc this is based on a in a mathematical formula we develop uh so the this formula has you know kind of a lot of input that this is not really the place to kind of expand on that uh but the thing is that like all the inputs here you know are kind of available to fetch from the blockchain

not necessarily from a smart contract but are available to objectively you know read from the smart contract via events or other means and it has one additional but only one parameter which is totally subjective and then given this subjective parameter we can also you know kind of compare the risk of different uh different platforms uh so this is what we did and you know well and how this thing could be used so right now standing Market spend a lot of money effort but also time on discussions you know in our discussions on you know

deciding what is the loan to value ratio and everything so like the vision is to make this process fully automated but even for the medium term you know the dial code provides some templates uh you know you know like on his vision and his confidence factor for these things and then we will not end up you know it's like with debate and discussion and typically the debate and you know the conflicts typically happen and you know like very they are very time critical they happen in a scenario where it's kind of the worst time to discuss these things because it's under a lot of pressure and you don't want to look like political but you know you also want to save your protocol so now the Dao can kind of decide on it prehand you know beforehand

uh and you know the actual decisions will be taken by a smart contract without any political bias uh also listing assets is a very important role of Landing markets so here you know we are giving this objective criterias uh adult could be more automatically be able to list assets you know looking at this risk Oracle that will provide you know like the quantitative data on the different assets and yeah you know to begin to put it in a bigger scope right now risk management is very political process

it comes with a lot of manual operations and actually manual operations were known to have also bugs like configuration bugs because we wanted to set one long to Value ratio you forget it to add few zeros there and users got liquidated for no good reason and you know actually lost money it's very governance heavy Boston you know kind of human emotion and you know it's very also very hard to get like a human no you know like to get attraction to community discussions but also it actually takes time and every kind of decision you want to make could very easily take like a month uh you know and finally it's a pack right now it's mostly dictated by risk vendors

who use closed Source models and then just come to the Dao and say look this is our recommendation and they can vaguely explain you know how they came to it but you know like there's no real way like to fully understand it Things become more neutral smart contact makes the decision it's fully automated so it's also saved a lot of effort and also resources and of course it is transparent and decentralized as device supposed to be uh so yeah so if you want to learn more we don't have time for a question but you know either catch me off stage and you know you can follow us on Twitter reach me out on telegram we have a

website for be protocol and Forestdale and yeah thank you very much [Applause]