Blockchain-enabled Data Marketplaces: Democratizing Data Access / Saul Martin

Transcript

mic check okay it works awesome so my name is Seoul Martin today we're going to talk about uh data marketplaces on chain um uh I'm a blockchain data engineer at nethermine within the Enterprise team um oh okay now Mike's working better um awesome so but why this topic um think in in light of recent events uh that we've seen so far uh there's racing concerns about the use and control of data that some parties are doing right now on that extent here we can see two different headlines from last week and they're a bit updated and I'm not gonna lie and with Twitter just reduced the previously available API for researchers

ready it has done some real stuff um with third party and developers actually for Reddit I just read before that there's going to be a blackout um tomorrow um for I think two days in which separators are going to be private because of the usage of the data and the newest policies implemented there so yeah I think but why blockchain so I think overall blockchain by creating an efficient and transparent Marketplace we can really help streamline their access and make it available so that we can make available individuals can access data and benefit from third data resources and potentially and hopefully also contribute to them so in this

presentation we'll talk about um first the benefits um on on a Marketplace on chain data Marketplace the challenges the disadvantages which I'd like to say there are also some challenges that can be dealt with and some real work examples so I I always say I really don't like to say this but like data is gold not the new gold but actually super valuable um brilliant to use this expression but it really helps to contextualize and put into situation uh the value of the data it has however that said the actual processes to harness the real value of data remained rather undeveloped and

actually the models remain relatively relatively centralized so um on chain there are marketplaces has a large large amount of benefits over traditional ones and there's lastly a bulb around the blockchain benefits so one we have transparency security and ownership on the transparency side everything is public so blockchain by nature is a public or historical Ledger in which all transactions are public verifiable and immutable which really can help make it so that both Partners or data Sellers and data buyers can really trust the origin of the data and the quality by itself then on the

security side blockchain again by definition has some cryptography features which make it that is can be protected against damprin and finally ownership so you can really give back to the owners data owners can really have that flexibility uh to essentially do as they wish for the data meaning they can set up their fair price or not if they choose not to do so and on the other side they can also help control who can access their data and into which form there's anyone that is not just put there but that one is about and around uh the new new business models and industries could emerge uh by this for

example small businesses and individuals can benefit uh and can actually uh create new um Industries could emerge by just doing the the analyzing the data that is actually uh placed on these marketplaces and one example of that is if you're a medical engineer or medical imaging PhD student and you're trying to get analysis of some tumors you don't have many data sets available right now if you could serve those x-rays Anonymous of course then you could have your models be richer and therefore can benefit from that wealth um cool so now let's go through the challenges so as I mentioned before

small business individuals have it more difficult why is that that's generally because um these entities there's a lack of transparency uh mainly due to many players so for example I don't know exactly where to access data for uh log level transactions of credit card users in the US that's super valuable data set but that's something that is not available right now so how can we deal with it so on that end it's kind of a challenging environment to step in so blockchains on team that have marketplaces actually have it difficult as well so uh first

one of the first challenges for pricing so there's the best amount of data available passive amount of data sets so it can really become difficult to fire to find a fair price and find the market for it then data quality that's the risk of low quality or false data entering which could undermine their trustworthiness of the marketplace itself I think here it should be noted that using an arbitration mechanism like such as chain link for example which Rewards or nodes if they provide accurate information or penalizes them if they do not then put could be quite valuable although it's slightly different as for example for chain that you might have 10 sources of data for

the same price for the same blockchain here if you're selling your own x-ray this might be only one source of data so that becomes tricky and of course we have privacy so with increase they are saying this increased risk to privacy so we have to make sure that user and animation is something that can be done easily and then we have the right consensus mechanism so that if the data uh needs to be dealt with and in in in in line with gdpi laws and whatever all the privacy laws we have then that can be done then let's look at these advantages I think uh briefly there's lack of liquidity so it might be difficult for

you to find a buyer for your product or you as a seller or you as a buyer find a data set that you're looking for then High fees so blockchains or actually data by itself requires a lot of computing um computing power which can make it difficult to to deal with in blockchain and then there's a technical uplift so not this is not something for everyone and I think this is something that needs to be tackled within the whole ecosystem by itself it's not only for data marketplaces so now let's look at some on-chained data marketplaces live examples so we have two right now so the first one would be uh the mobiles and second one

will be also and protocol so T-Mobile stands for decentralized marketplaces of online big analytics this is actually a resource paper by tennis University that was published I think five months ago so they essentially I'll go through the details later and then we have ocean protocol as a real world example so this is something that is in production and you can actually uh interact with so this image might be uh some might look a bit cumbersome and it is so let's go through the steps that theoretically on this theoretical approach uh a data Marketplace would work based on these researches so first um let's say I'm a data owner I want to

sell a piece of data the things first thing I would do is choose upload this data to DSN or distributed stored Network once that's done smart contracts will actually check if you have uploaded the data to the specific location how does it work because you actually instead of you uploading the data to blockchain you upload the data access keys to the blockchain that then is ciphered and is encrypted so first step you upload it second step the smart contract checks that you have actually uploaded and is something that is reachable then you can return the ciphertext for that

so now let's say I'm the other persona I want to buy that same piece of data how does it work first you would need to find a place to actually look for the data this is something that hasn't been specified on the paper itself ideally you have a UI in which you can choose search but what it did it does mention is that they can use ipfs search capabilities to index the hash or kind of the data set itself and the metadata and then you'll be able to um research and like look for the specific data set they're looking for so once it's indexed I'm as a data buyer they can go through and buy that same data set how this is just how does it work so I'll just go there pay my fees pay the

price that has been set by their owner and I will return it both smart contract will return decipher checks meaning I will just get some encrypted text with data access keys to that data at the same time the owner will return to the data buyer be a TLS so secure mechanism the decryption key so that Cipher checks or the access with that the access key to that data therefore you can actually then have access to encrypted text and have access to the word that data has been stored currently so what happens if there's um a bit of abuse by either the data owner or the dead by year so they have said an

arbitration contracts in which if um necessary or if something has gone wrong through the transduction in terms of data quality uh data buyer can go and ask well can I actually trigger the smart contract and then trigger a whole arbitration mechanism which I think it's like five seven nodes that need to validate if the data is actually what it used to do or like yeah so that's on the theoretical side on the actual implementation done from awesome protocol there's two ways I've divided this ocean protocol kind of approach so you could either serve public data sets or you could share and

access private data sets so how do we get access to public data sets and these might include comments um gradients from decentralized apps for example so first the way it works is that awesome empirical tokenizes the data access Keys these will be called the data token and there are an rc20 token so once you want to access a piece of data let's say the historical eat prices you could you'd be able to buy that for the price um if there's a prize and if there's no prize we'll still have to pay the fees to get it you will retrieve the access keys for the data is located and then you can run whatever ml AI analytics um just statistics that you want to do or

just keep the data if you want to resell that as a service for example you're running I don't know a risk indicator for a financial service and you have as an input the historical ease prices and it's an output an indicator you'll be able to do so by just meaning a new nft and then that would work as a sub licensing mechanism also something I forgot to mention is that this data is actually you're also your only certain data access keys but if the data set is small enough you could theoretically store the data within the um kind of the nft itself so what happens if you want to sort of private visit this is I guess best use case so that case what you would do is

you have still your um data token but you have as well a new erc20 uh token that will be an algorithm what is an algorithm that'll be any piece of code any python code Java whatever you want to do you put all of those together within the environment they call it computer data and that runs in a Docker container and you will get the output or the logs of that processing job um I think this might be something that might come up in a quick q a is that what happens if my algorithm just prints the whole data set um I think that's going to be quite an abuse pattern that I could see quite frequent here

right now there's not an average Asian mechanism on also protocol however the data owners can actually whitelist a set of algorithms themselves so that you know you can bet manually if that algorithm is actually a good one and then it's not really abusing your data so then let's go through real world applications like why do we what would something like this take take take importance so of course we have Healthcare as I mentioned before if you want to do some Medical Imaging research paper you would need some data to actually

um to actually get some output so if you want to sell your X-rays I don't know whatever medical processes the the can be done you could eventually do so on the blockchain and then you could get paid by doing that Financial Industries mentioned before briefly this application like prices are most probably public already of course it depends on the Fidelity and the granularity and all of that the instruments that has the cost but I think that the the option here from the financial industry is that they will be able to have customized risk indicators which you can sell directly on non-chained data Marketplace and then of course we

have the environment so actually think Catalonia and Barcelona have done a proof of concept of this so you will be able to monitor the CO2 emissions or on a given City and for this it's similar to healthcare so let's say you have and you put a device in your house to monitor CO2 emissions or to monitor um I don't know the temperature you could tell that if you do that right now you potentially don't have any clients because one data point is not really worth it but if you have a neighbor full neighborhood doing that or a full City full country doing that then that becomes more valuable because you have more data points and then as a conclusion

these marketplaces exist they're a real thing however they can be quite challenging to be implemented mainly because of private Solutions and and also they're not fully decentralized right now as you're kind of dependent on where you store the data as you're actually only sending or storing the data and giving them the access keys to the data so that could be stored in honestly bucket could be stored ipfs could restore and file and could restore anywhere so that's not really decentralized for some options as I mentioned before um but yeah I think the audience has any questions be happy to to answer them foreign thank you so much does anyone have any

questions hey uh thanks for the talk um I was wondering if you could maybe elaborate a bit more on the data aggregation part so as you were saying like if you are um an individual measuring your CO2 in in one um household it's not very efficient but you could aggregate this data from multiple sources is there already a sort of decentralized infrastructure that that helps with this aggregation so that it's not a cumbersome process where I have to contact all these different households to see if they want to give me my data yeah I guess for the environment data this there will be a bunch of different

options this is more broadly speaking how that will work perfect sample I can think of a Marketplace in a Marketplace that could potentially use labels as a filtering mechanism so let's say they have an upload your data kind of drop down on their website then you would select the data you want to upload CO2 emissions for example and then that would be always kind of aggregated so if you're a buyer of that data let's say you're working for the government trying to compute I don't know flat predicting floods for example although it's not really related to CO2 but yes um they will just be able to buy the whole kind of data set for like all the individuals that have actually uploaded

the data by a single click so yeah it should be possible to do so it's just um then the actual adoption right now is not there yet so um it's it's a bit difficult to do right now um how how do you prevent like node operators from doing like a vampire attack because once you have information you can't unknow it and so you could have a a competitor protocol that does a vampire attack and just like collects its own data and collects its own yeah I think that's uh that's a risk um here because once you have the data you have it

um so oh it's not here yet but anyways I search for example so one extreme private data sets another one is certain public data sets for public one we don't care for private data sets what's been done is that you put your algorithm together with your data and then that runs separately and something to note here is that these servers are actually centralized that they're using AWS or S here to run those kind of compute jobs on the cloud so it's kind of susceptible to an attack there theoretically if you bet the right algorithm you would only get locks that this algorithm is supposed to get you so yes it's kind of tricky how to do that that's why privacy and setting up

all of that infrastructure is kind of valuable thank you thank you [Music]