The Black Box of Centralized Exchanges

Transcript

foreign hello everybody my name is Clara medalli I am the director of research at kaiko this is my third year presenting at ECC so I'm very happy to be back again today I'm going to talk about the black box of centralized exchanges so paradoxically over the years transactions on blockchain using blockchain technology have become less and less private this is due to the rise of on-chain investigators whether that's just your average individual on Twitter or entire companies whose purpose is to tie

real-world identities to on-chain addresses however the one exception to this rule is centralized exchanges this is because all transaction data on a centralized exchange is anonymous which means it can be very hard to understand what's actually going on inside so today I'm going to give some examples of how despite its anonymity this data can still be used for very powerful insights but first a little bit about kaiko kaiko is a crypto data provider we were founded in 2014 and today we have four main Services the first is our Market data this is centralized and decentralized exchange data from more than 100 separate platforms we also offer BMR compliant rates and indices If

You're Building investable Products we also have quantitative analytics for pricing and assessing risk and then finally research you can subscribe to our newsletters it's all free online research.cico.com so I think it goes without saying that centralized platforms have really harmed trust in the cryptocurrency industry over the past year whether that's FTX or Celsius it's these centralized platforms that have both enabled bad activity and bad behavior and also been the perpetrators of bad activity um so the question I want to explore is centralized exchanges are they truly The Last Haven of privacy um crypto is increasingly less of an effective tool for Crime simply because

all transactions now leave a permanent Trace you have law enforcement that are getting way better at tracing this you have a ton of blockchain analytics firms whose core goal is to identify and Trace bad behavior on the blockchain but again centralized exchanges it's really hard to detect Market manipulation happening on these platforms because ultimately it's the responsibility of the exchange to do so this is because all of the data is anonymous whether that's trade data or order book data there are no identifying information to the individual when you're looking at market trends and what's happening on these platforms and I think that that's ultimately after the past year that's not good simply because you see what

happens when you have these centralized platforms unchecked like FTX where essentially you had the trading firms trading against the clients on the exchange and a bunch of other bad behavior but it's ultimately still possible to gain insights so today I'm going to give three examples of the types of insights you can derive from Anonymous centralized Exchange Market data so the first example I want to want to explore is looking at wash trading so wash trading is a type of Market manipulation that involves essentially a Trader trading against themselves back and forth the goal is really to inflate volumes whether that's to show that an exchange has high volumes or a cryptocurrency has high volumes and the analysis I'm going to show the goal is

not to identify individuals here rather it's to show how exchanges it's their responsibility ultimately to employ Market manipulation detection so watch trading ultimately Still Remains a pretty big problem in crypto thankfully it's not nearly as big as it was in 2018 this was during the Ico boom where you had many many uh small cryptocurrencies that were launching and there was a lot of bad behavior happening on these unregulated platforms um and ultimately wash trading is enabled by exchanges that have lacks Market monitoring if the exchange is not monitoring this then there's no reason why an individual won't just go on and trade against themselves if it incentivizes their project in fact in

the recent SEC lawsuit binance us was just accused of enabling lost trading on the platform so this is to show that even today it's still a big problem even big brand name exchanges still often do not have the correct Market manipulation detection measures in place not only that watch trading is incentivized by these big price display websites which often rank exchanges by trade volume so if you're let's say a retail Trader that doesn't know much about the reputation of various platforms you might go on these websites and rank the exchange by volume and you'll notice a bunch of exchanges that seem like they have huge volume but at the end of the day you haven't ever really heard of them but this is impossible to know if you're new

to the industry so at kaiko we wanted to better understand how can you actually detect uh wash trading so it's very complex but on the spectrum of sort of easy wash trading detection to complicated wash training detection I'm today going to show a very easy example but again the more sophisticated you are the harder it is to detect so here we did what we're calling the volume to Market depth ratio what we essentially this is the assumption that exchanges that have very high trade volume also have very deep order books so you should see a pretty low ratio when you're comparing trade volume to order book depth and so exchanges that have a high ratio they suggest that they

have inordinately high trade volume while they have very thin order books which doesn't really make sense from a Market's perspective so what we did was rank 20 cryptocurrency exchanges all centralized using the volume to Market depth ratio and we can immediately notice two clear outliers the first in his exchange called bitforex It's relatively unknown yet it ranks extremely highly on platforms like coin market cap when you're ranking by trade volume the second exchange is quite known it's binance but there's one specific reason why it has a very high ratio this is because binance had a lot of zero feed trading pairs so zero fee trading pairs are quite controversial simply because when you have no fees

there's no reason why a Trader won't just trade back and forth especially if there's no measures to stop this but we're not going to focus on binance today for this example let's actually dive into the tick level data on bit4x so when we were exploring bit Forex trade data we looked at many many different trading pairs we also compared The Tick level data to more regulatory compliant exchanges like coinbase just to make sure that we weren't identifying any outliers and we noticed big differences in how the tick level transactions looked so on this chart we charted all tick level trades over the course of a week for the LTC Market this is Litecoin the cells are the red crosses and the green circles are a buys

and when you see a Red Cross in a green circle overlap into a Crosshair this is very clearly what we think is probably a wash trade simply because the trade was both a buy and a cell executed at the exact same second in time for the exact same amount this is not normal Market behavior and in fact we noticed many many crosshairs just over this one week time period when we compared it to coinbase we did not see this when we compared it to Kraken we did not see this which suggests that this is Market that this is not Natural Market Behavior so again I'll bring up binance us simply because they were just listed in an SEC lawsuit alleging uh alleging that they had no trade surveillance or

manipulative trading controls until February of 2022 not only that the lawsuit alleges that the exchange's matching engine allowed users to trade with themselves so this is a huge problem simply because this means that before February 2022 you can't really trust the market data how are we supposed to know that the trade data that we were getting was not influenced or at least partially due to Traders just trading back and forth against themselves so it's very very important that exchanges have proper Market monitoring in place because ultimately this is a form of Market manipulation so the next example I'm going to give is U.S trading activity this is not Market manipulation however we live in a new

reality in the crypto industry where Regulators in the United States are very much cracking down not only that you have many many centralized platforms that have tried to evade the rules against Americans using the platforms and this has caused big problems we have multiple lawsuits by both the cftc and Securities and Exchange Commission that are essentially accusing centralized platforms of enabling Americans to use the exchange and this is not any opinion that ultimately crypto is all about open finance but we do live in a regulatory reality where the United States is very much cracking down so let's again look at the example of binance simply because they were listed in the recent lawsuits so here what I'm charting is hourly trade volume and in

Orange you see U.S trading hours this is UTC time and you can notice that three weeks before the cftc lawsuit earlier this spring we can see that hourly trade volume was very much concentrated during U.S trading hours this is on binance global and binance derivatives exchange which Americans are not allowed to use and we can instantly see that after the lawsuit was filed this trend line has very much evened out we now see much more even trade volumes across the various hours of the day not so concentrated during U.S trading hours this suggests that there were instant changes either at the exchange level or the traders that were using the platform but what's interesting is if we go back

in time let's now look at 2018 and compare it to 2023 here we're looking at this time binance which Americans are not allowed to use and coinbase which is a us-based exchange and we can see that in 2018 binance trade volume was quite flat over the hours but by 2023 this volume was heavily heavily concentrated during U.S trading hours not only that it was more it was almost the exact same level as coinbase which is an American Exchange the vast majority of coinbase's clients are United States um so this just shows that there clearly were well it's not definitive evidence there was a lot of additional evidence used in the lawsuits including internal messaging and examples of non of American clients they were letting use

the platform but this is to show that just by looking at the market data we can already pretty much see this trend so the final example I'm going to give is on decentralized exchanges Regulators have not yet looked at defy but we also need to be very conscious of the fact that that doesn't mean they won't and in fact when you just look at Uni swap V3 trade volume this is for one of their most liquid liquidity pools we noticed that in 2023 hourly volume is very heavily concentrated during U.S trading hours when Americans are awake so this doesn't mean anything there hasn't been any actions but it's just to show that today the vast majority of crypto activity tends to be concentrated during U.S trading hours

so the third example I'm going to look at is a bit more complex this is an oracle exploit so Oracle exploits rely on both on-chain data and off-chain data and so I'm going to show how you can sort of use these two data types in conjunction in conjunction to gain a deeper understanding of an oracle exploit so first what is an oracle exploit today it's a pretty pervasive problem for Lending protocols it's getting a bit better but ultimately what happens is any D5 protocol that leverages a price Oracle the price that they source that they source is typically from centralized exchanges for example altcoin markets on centralized exchanges they're often used for Price oracles and

in fact if you manipulate the the altcoin market on a centralized exchange and let's say the pre you make the price crash this can flow through the Oracle which in turn can trigger liquidations on a D5 protocol so that's what an oracle exploit does it manipulates the price feed in order to financially benefit the exploiter and this has happened numerous times over the past year for millions and millions of dollars in fact it's caused a lot of lending protocols like Ave to Halt lending pools for Less liquid altcoins it became such a problem um so let's give one example this was on Ave which is the largest lending protocol this was an attempted exploit by Avi Eisenberg the notorious Oracle

exploiter back in November this was actually a failed exploit so he tried to profit off of the CRV token But ultimately failed to do so but he are the same exploiter had other successful attempts um so here what we're doing is using kaiko's on-chain data to chart every single transaction that Avi did on Ave we can see the borrows the deposits and ultimately the liquidations which was the failed attempt of the exploit but it's quite complicated all of the steps that he took so I'm not going to go into detail explaining the exploit what I'm most interested about is what happened on the centralized exchange because a lot of the Oracle exploits that I see in the analyzes they're mostly looking at

the on-chain stuff and that's just because centralized Exchange Trade data it's less public and harder to get a hold of but kaiko has both on chain and off-chain data so what happened with this specific exploit is we saw on chain that Avi transferred CRV which is curves native token to okx okx is one of the largest centralized exchanges and he instantly started selling this token and we can see this in the data we see that the price of CRV started to collapse um and trade volume started to Skyrocket this is very clearly an exploit attempt because there was no similar Market activity happening on other exchanges so ultimately this exploit failed but it shows the power of both off chain and

on-chain data so ultimately I showed three examples of how you can use anonymous centralized exchange data to gain deeper insights into bad behavior the goal is not to expose individuals rather be exchanges and to encourage better uh Market integrity so before I finish kaiko will soon be launching a data visualization platform in Q3 and it will allow you to monitor liquidity for many crypto assets at the exchange level so very granular trade data and order book data to learn more about it you can subscribe to kaiko's research newsletter we publish twice a week we have really in-depth analysis it's free and you'll hear about all of

our upcoming product releases which are more targeted towards a wider audience rather than just Enterprises so thank you very much and looking forward to seeing your subscriptions foreign