Unveiling the Future of Zero-Knowledge Proofs - TOKEN2049 Dubai 2024

oh hi everyone oh okay so uh so today I will be talking about live poly partnership with Google Cloud on live doing the GE knowledge proof Cloud so uh I think I will just start my slides okay so poitra is actually a very big user of l z knowledge Pro and also a very big user of like Google Cloud so that like since we switch from AWS to Google Cloud we are kind of spending about 100K per month and what is this 100K go to we are actually using that to generate like zon not proof for our ZK Bridge so this is actually also an Evidence about the demand of light zon not proof why do we need to generate this many l zon Pro because like in the past like six months we have generally about 1,000 like cost Bridge like cost change transaction for like a lot the user so this is kind of like the cost for you to serve like 1,000 like one one like 1009 million that users so one of the reason that we use Google cloud is that like it actually provides some like very good like service guarantee that we need for our user first of all like we need to be able to scale up and down because that like user don't always like it's not the same every day and and that like sometime because you have some agile compan that will be more user and sometime like when you when the Chinese people we up actually they are also going to be more model that's our like like observation another thing is that we need to be reliable because that that user actually want to cross bridge at different CLS of the day and if the the to cannot be generated on clamp the transaction they started in the middle user will be very worrying user will Panic if the money is already lost so we have been using gr Cloud to do the pro generation and people will now ask this question is that okay you probably also have heard about different kinds of the decentralized solution for doing ZK Pro generation actually that have been a few company that are all talking about the central life Pro generation and that like one of the reason why we are thinking Google cloud is that Google cloud is actually not very different from other solution if you are talking about the pricing model then Google is just one that g you the same price regardless of the demand or the LI Li Vol of the network and another thing that we find very important for our use cases and for other use cases for ZK is that if you are using a decentralized platform to do proof generation you have the privacy and security issues why because when company actually want to use the cloud to do pro generation they probably don't want to open source their design they don't want to open source their approval design they don't want to open source like users data there are some application in ZK where users data cannot be public and for those like application you come into the situation that you need to use the cloud services where your security have some guarantee Google turn out to be a very good option like uh we are kind of Designing our Pro cloud in a very specific way search that like we are using Google cloud and we as the person that building this intermediate layer will not be able to see the program that you approving and also the data that you are proving over so this protect the users's privacy and also our clients like privacy okay so and people will probably be also asking that like oh is it a big problem that you are using Google Cloud to do pro generation isn't that becoming very like centralized so our argument is that like at this moment we don't think like the centralized like system are like really reliable enough to do this and we find that actually uh first of all you cannot be very completely decentralized Pro need to be powerful enough to do the proof and that like uh you you have ZK you don't actually need to trust the approver appr Dr can be anyone that can get the job done and that like today if you are looking at the ZK R up if you look at different kinds of the ZK company most of the ZK company actually accept Tao most of the DK company are actually using centralized Prov and one of the reason that we see is that maybe people will also say that you can use Other Des centralized solution they will be cheaper we find that actually if you want to lower the cost of pro generation it's not about finding the one with the lowest like hours per like like dollars per hour what you need to do is trying to find a way to optimize the machine alization so in our like internal like use of the Google Cloud we find one thing is that first of all is that Prov you cannot just span it out when you need to use it because that like actually so there are a lot of the engineering things about the Prov you need to let the code run in memory it will be faster it's actually highly performance sensitive to the machine that you are running it and that like uh but Prov don't usually don't always have time to do for example in our case as a bridge user comes and goes and in every day you actually don't have like 30 like like usage and you don't have all the machine busy all the time actually we are we are okay to admit that actually based on what we look from the graph the CPO ization is definitely below like 30% which means that that might be actually a lot the opportunity for us to really do more and what does this mean is that that might be actually a demand to build a Shar proof platform such that like different application will be together utilizing this machine to do the pro generation because you don't actually have enough utilization because of the w l because your applications okay so we are announcing like Po here just like Pro Cloud platform and this is basically our collaboration with Google Cloud we will be building something on top of Google Cloud such that like people different typs of dzk application can use like Po just like Pro Cloud to generate proofs okay so is based on Google Cloud uh you only need to trust Google Cloud to be functional you don't need to rely on us so basically we build really directly on top the Google cloud and if Google Cloud have machine you have machine to do the proof like so and you don't need to trust polyedra for like security we are we are working to design it in a way so Google have actually some sort of like unave things we are trying to use those tools such that like polyedra even if we are compromised we are not going to see the data that you are proving over so that your business squl even if you are competitor to us you you don't need to really worry about that and we also so another reason that we specifically want to do this is that like we my uh our software is actually quite useful to allow the company because like when people are generating the pro in ZK a lot often are not really production level you don't call into a situation where you need to really generate 100 Millions like Pro like in six months and we find that even if you you are using CES open source library and you want to use those library on a machine on a server machine that have a lot the CPU you are going to see that all these Library are not really designed for that they actually run slower than with a few call on those like B server so we are also open sourcing our like internal like appr that are actually very nicely in working with this like servers and like our team actually have been working on a few like proving system that actually have very good efficiency and we think that this is also a very good to share like the result to like everyone okay so the core idea of this proval cloud is that we want to make it functional you have aov generation class send it here it will get processed and this will be very similar to like Google Cloud function so it's basically a function you don't need to create a machine and you only need to use this service when you have one task to do and you will be able to offer you a price that actually only depending onl the proof generation cost you don't need to provide upfront like commitment this is very important for early stff company because so we have one example so uh one our co-founder once like open 100 machine and it was actually for paper to do experiment he forgot to close that overnight that is already several thousands dollar so for staff company if they run want to run big scale when they actually don't have that kind of the world is actually the best solution for them to use something that is a functional like cloud services so it will be a very very simple interface you just provide the program and the data and our cloud services which actually will be running over Google Cloud as long as Google cloud is running this service will be running it will actually be providing the proof to you and two thing important here is functional we don't need to we don't require user to provide reservation or commitment this is what better for like early startup company and also we don't have access to your program or the data this is also good for Star company that they want to not disclose their business secret and also is important for ZK application where user like SEC information shouldn't be shared to the public okay and we also will be like open sourcing like our software tool that is for like generating ZK proof uh this will be in May this year uh so we will be open sourcing our approver our verifier that have been deployed in so many like blockchain and like one of the compiler that we never talk about before that actually was very useful I will show that very quickly and we will also show some benchmarks that people can know what are actually the performance of the like system that are now currently used in production okay so we are going to present a compiler that like people who are already writing ZK circuits in G will be able to use this compiler and directly compile to our new protocol so existing startup company they can just quickly like migrate that existing design into the new one and this Pro is going to be much more efficient than the one that people have been using before it's really a art of like letting the compiler to find what are the common pieces in the ZK proof and try to put them together this will allow the proof generation to be running in a fast SL speed and so we want to make our inter we already are doing that basically that the interface will be similar to existing Pro generation tool people will be very easy to migrate and that like we support different types of the commitment scheme so that people can just use the one that they are using we are allow like very good selection of the like modern CL modern curve and like also the very recently there have been a lot of discussion on M31 and circle start which we are also going to support okay and this conclude my cloud and if you want to know more about ZK Prov cloud and other thing that like Poli are doing including Bitcoin and you can go to and follow our Twitter okay thanks everyone thank you waking but please maybe we still have time for questions so use this time if anyone would like to ask more in detail this is your chance so please raise your hand we'll give you the microphone and uh let's find out more okay thank you thank you for the speech uh do you have any specific area that you like and that you work with like uh digital Identity or maybe some specific area not only cloud in general uh uh so we are actually specifically you mean that what are the specific area of the cloud that we are working on right so for this one we're really specifically focusing on like ZK Pro generation like for this like ZK applic for ZK roll up or for ZK Bridges the reason is simple is that like you can see we are building over Google Cloud right so that we feel that like Google will be basically our Hardware layer providing all the infrastructure and we are more position ourself as basically the software layer like basically that like we are just contributing the software that we have that is not available outside and that like we kind trying to make it easy for like Star company to use that and you can see the functional pice also for that because we can see a lot the like stop company struggling to kind of like run that approver like even setting up the machine to run that is very difficult but we feel that like well just join us we are already doing proof just come here and that I we do that together yeah okay yeah