Proving Ethereum Full Nodes in ZK - TOKEN2049 Singapore 2023

hi everyone it's really nice you guys are here and I'm Tian Tunisia from polyhedron Network and I'm co-founder and CTO and today I'm going to present the proving for node of Israel in the Eternal proofs and we are going to present more stuff here so Bridges is our main product and it's a risk it's very risky one because as we have witnessed a lot of bridge accidents over 1.5 million assets has been lost in last five last several years and to mitigate this vulnerabilities we have proposed a Decay Bridge it's based on a zerology proof and using large proof we can minimize the trust and we can much simplify the smart contract development on chain and more exciting thing is that we have delivered more than millions of messages since we launched so the core mechanism of the ZK bridge is we use 200 proof to prove the consensus and we have integrated a lot of different chains there one and there too and this diagram is uh it's a framework of JK bridge and as you as you can see here we have two blockchains don't know that on on the purple side is Center on the blue side is a receiver and our proof network will monitor the center chain status and it will prove the consensus of the sender chain using geologic proofs to be more specific let's say sender chain is a serum then the proving task will be you prove uh thousands of signatures from radiators that has signed this block and super majority of this radiator has agreed on this block after the block is proved it will be sent to the receiver side for verification and if a receiver is convinced that the block is actually correct it will store it and user can retrieve their transaction from this block and do more operations on the receiver chain so to enable this proof of consensus we need to innovate on the proof system so today I'm going to present a next generation of Technology proof it is one of the it is actually the fastest technology proof system and here is the structure of the proof system so we use the basic component algorithm it's called Virgo and it's based on the gkr protocol it is extremely fast that can process millions of minutes of constraints on a single CPU core and in the Virgo we combine all of this processing power in one proof system and we can prove where is gigantic circuit in this proof system and furthermore we will use recursive proof to compress the proof size so that the proof can be efficiently verified on chain and this proof system is the world's most advanced proof system in a single CPU core it can process like three millions of constraints per bit per second and it requires extremely low memory and if you combine all of this CPU cores together you can prove billions of circuits a billions of sites in terms of circuits so let's look at the serum specifically to prove a full node consensus we need to verify that at least twenty thousand validate her signatures and these validators will rotate frequently and the time limit is actually 12 seconds because insurance will generate blocks every 12 seconds if you compare it to a growth system growth is still need more than half an hour to generate such proofs and our proof system is much much faster than 416 this is our proof system both of the line are put our food system the the blue one is we run our system on a giant server which is which has two AMD most advanced CPU installed and the purple one is we run our proof system of 64 smaller device and this is much faster because we have a lot more CPU cores and to give you some idea the proof generation time will be like five to six seconds on our smaller device and the energy conception consumption is kind of negligible per proof so people may talk about decentralization of proofs decentralization of proof is extremely important because we want incident we want to encourage everyone to generate 300 proofs the major roadblock of generating theology proofs is that usually their approved requires a huge machine that costs more than 10 000 US dollar this is a huge uh upfront cost for average Miner or validator so in our proof system we reduce it to a very low spec machine and I will present it in our deck here here is our machine this is a gaming console developed by the wealth Corporation and you some of you may know that they have a great gaming platform it's called Steam and this is a steam tag this machine costs about 400 dollars and you can you can purchase a lot of them from the steam there from Steam yeah and to for validator to become a gkp brewer you only need one of this machine and to finish our insurance around your proof we need 64 of them but it's okay that they are they can be globally distributed the communication cost is actually no so this is kind of a pool of zero improved machines in summary we have very low Hardware requirement the only thing we require is you have a stable internet connection we don't want you to drop off during the proof generation and the CPU and the ram spec is a requirement is super low so next we I'm going to do a demo and yeah please switch to my deck here's my steam deck and I bring one here of course I cannot bring 64 of them so this will be a one machine demo and first let me show you our let me show you the sting it's back so we are running the approver on Steam OS and the hardware is we have eight core CP AMD Apu this is very cheap Apu by the way and RAM consumption is this six volts slower lower than 60 16 gigabyte of RAM and to prove that I'm not I'm not using a video I'm going to go to either scan refresh it foreign [Music] it's really hard to navigate using steam deck yeah here's the latest ball hash you guys can check I'm not I'm not recording a video and I'm playing a video here you can check this like 10 seconds ago okay now I'm going to go back to the to the console here so I'm going to run a signature verification algorithm signature verification back proof generation algorithm on 256 signatures and the eight is actually eight CPU cores this machine has 80 CPU cores so I write 8 here okay it's proving it layer by layer so since it's a gkl algorithm and as you can see we proved in three seconds and if you're going to look at more not your circuit for example uh here is a Circuit of size 1000 and sorry 1024 let me show you here 's a circuit of size 174 and the time will scale linearly it should be like around 10 seconds okay like 12 sec yeah so if you are targeting like a silver and blockchain you could use this 124 configuration and you will need like 32 steam decks to finish the proof yes finish the demo yeah thank you so much so with this proving power we can actually do a lot more and we are trying to propose that proving all uh we are trying to propose to a decision that we can actually put all validators into one slot and this will achieve very fast finality and improves scalability let me show you why so in current ethereum solution we have around 1 million validators and here is a smaller scale illustration uh assuming that this is one million validators and in current ethernet solution they will divide this one million validators into 32 groups and here we are dividing into four groups for demonstration and each group is named as slot so they will have like 32 slots for for epoch and if you want to a slot is a block actually so slot is equal equal to a block if you want to achieve like confirmation you will need super majority of validators assign your transaction which means that in in this diagram you need three blocks to approve your transaction so you need to wait for three blocks and if you are in real estate you need like 20 blocks to confirm your transaction so the confirmation time is actually several minutes and why that does they do this design because they their P2P Network has overloaded so here's a example of PTV network if you have a very small slot then your P2P may be okay the the load of this p2b network may be okay but if you extend to the whole very data set with 1 million radiators your P2P network will be super messy and they cannot handle it so with a theory proof we can do the following we instead of sending uh signatures or cost of instances in the network we prove the signature are correct so actually the validators only want to know that their neighbors whether their neighbors are honestly signed it or not so we don't provide directly we don't directly provide signatures in a two degree network instead the validators generate technology groups and proofs is extremely short so they can communicate more efficient and this P2P network will not overload so we can put every validator in a single slot and with this we can scale we can basically confirm every transaction in one slot so you only need to wait like 12 seconds to confirm your transaction and for timeline we are doing our CPU implementation of approver and the efficient distribution care has been improved has been implemented as a demo here and next next several months we are going to do a mid and large scale testing and for next step we are going to schedule a GPU implementation of course and compiler implementation for our circuit and currently we are using gnoc which is a great compiler and this compiler is super stable and super efficient their front-end language is super easy to use so we recommend you guys if you are interested in developing CK circuit you can try Gene lock and we will also build our compiler from their front end and we are open to discuss some ideas on optimizations I think this concludes my cloud my talk thank you so much and yeah [Applause]