# Jehyuk Jang - ZKP for Everyone: Apply ZKP to Your Ethereum Transactions in a Few Clicks

- Channel: [ETHCluj Meetup](https://streameth.org/ethcluj-meetup)
- Date: 2025-10-07
- Duration: 18:29
- Watch: https://streameth.org/watch/yt-bxaggBMRzco
- YouTube: https://www.youtube.com/watch?v=bxaggBMRzco

## Description

We present a new approach to zero-knowledge proofs (ZKPs) with a focus on usability and efficiency, grounded in our prior research (eprint.iacr.org/2024/507) and implementation (github.com/tokamak-network/Tokamak-zk-EVM).

Our work addresses the challenge of transforming Ethereum block verification into ZKP proofs. Existing solutions—such as zk-EVMs and zk-VMs—typically require deep technical expertise or access to high-performance hardware.

Over the past two years, we have been designing a more accessible alternative. The system we introduce in this talk requires no specialized knowledge and runs on standard personal computers. This enables anyone to deploy and operate their own ZKP-based Layer 2 network- easily and affordably.

## Transcript

Uh thank you for coming for my talk here. So yeah, as introduced I'm uh Jay-Hawk Jang. So and you can just call me Jake and I'm from Tokang Network. Uh today I'd like to talk about the TKP for everyone. And we will explore how to apply GKP to your some random transactions, Ethereum transactions in a few clicks. Yeah. Yeah, here's what uh we will cover today. So it uh first uh I'd like to briefly introduce the techn technical things uh to uh general analysis proofs and then uh some key con considerations when designing your GKP and finally uh we will propose our new approach for easier and faster GKP utilization. Right. uh when you when you Google GKP uh you may find some many technicals uh sorry articles uh that praise that GKPs are magic kind of magic. So in my opinion uh no they are not magic they have clear limitation in theoretical some theoretical limitation and their limit stays uh within the bounds of the classical information theory which was established in 1940 by cloud chon and also in practice when it comes to uh applications to the etherium the GKPS are ju just just reshaping the scalarity tile uh this triangle and they don't expand this triangle they can't expand it they just reshape they they are just tools so uh yeah so I'm going to deep inside get into the uh GKPS and to this end I think I need to take an example of the use case for the GKP. So I think the the best best use case one of the best use cases for GKP can be said as uh verifiable computation. So in this scenario let's say there is a computation a big and huge and massive complex computation and you are going to outsource you don't want to uh do this yourself instead you are going to outsource this computation to a worker or prover so in this in this scenario uh the worker or prover will uh give back you the output put output of the computation along with the proof proof about the correct correct execution of this computation. Yeah. And so the key point here is how to make the proof how can you guarantee that the proof uh guarantees the correct execution of the uh computation. So I'd like to say the the entire procedure of making the proof in three steps like uh compile, commit and prove. So uh first first we need to compile the computation itself uh to generate the circuit. So here the circuit can be said as a list of constraints and with this constraint we can check that if the computation was correct or or not. Uh and given this circuit uh we need to commit the result of this uh compilation. So the commit commit step uh commit to the circuit. So the commit months the resulting commit months will be uh a comp the uh compressed representation of the circuit and also these commitments uh cryptographic commitments are some kind of cryptographic agreement on the on the integrity of this circuit between the prover and verifier. And then finally the prover the prover can uh generate the proof and this proof can be verified by a verification which comes from the commitments. Yeah, this is the entire procedure and here we might have uh there are some cons security considerations we need to take into when when we designing or utilizing a GKP. Uh the security potential security leaks can come from the first two steps commit compile and commitment committing because the the fundamental theories behind the GKP system does not guarantee does not cover this the security of the compile and commit. So they are out of the GKP coverage. So we need to take take we need to be very careful when when you when you compile and commit. So so I'd like to say uh there is a trade-off between the security and efficiency when you using when you're using such GKP. So depending on the use usage of the compilers and commit months uh if the system has more frequent use of more more frequently uses uh compilers and commit months uh that leads to faster that may lead to faster proof generation and easier privacy. But the system the GKP system will relies on more more will will require more reliance on the trust in the compilers and conversely if we use less frequent use of the compilers and commandments the slower the proof generation will be slower but uh the required the trust require uh dependency in the compilers will be uh less. So this can be a tradeoff when you're using the GKP and so this is the main topic today. So if you just want to apply GKP to uh any random transactions uh do you need to take all those things into account? So I'd like to say no here. So uh at this point I I'm going to propose our uh new tokama new GKP system called tokama GKP which has some good properties like this uh is to you which has easy to use uh graphic user interface and also has fast proof generation and has low memory requirement and cheap verification as cheap as other snacks. So uh with with our tokP you don't need to uh build circuits on your own. So you're you're free from compiling some things and we have we have you can generate a proof for any random transaction just in few minutes on your home desktop. Right? Then I'd like to say what make what makes this efficiency. So uh I'm going to compare our system with some traditional approaches like GKEM uh designed by the PSE and scores they are famous and yeah in their approach they compile the EVM the entire EVM into a massive circuit and this circuit should be massive and large enough so that it can prepare it prepare for all uncertainty and unpredictable computations that can be handled by the EBM. That's why this circuit uh cannot help avoiding the the size expansion. But in our case uh we adopted some two-step compilation approach h to reduce the uncertainty involved in the circuit. So uh we have two uh compilation approach steps. For the first compilation we uh compile the EVM especially not the not the full EVM but just static parts of the EVM that that is very predictable. So because there are no uh dynam dynamic things or unpredictable things in the circuit the the resulting subircuits uh must be very small and once you once you you get some contract and method to of of interest uh you can compile it once again so that uh we will generate the wiring of these subcircuits that can represent the this contract the documentation in this contract and method. So this is our approach and right so here I just summarize the the input and output of the tok. So as a result of the first compilation uh we can get some EV the commitments to the EBM subircuits which is called CRS and this CRS can be reusable for various contracts and computations. And next for the second compilation uh uh we can get some commitments to the wire subcircuit wiring which is called BP. And this BP is reusable for some various instances as long as as long as the contract method is fixed. That means if we want some want to generate some transactions coming from the other contracts then we need to rerun Yeah. All right. And here I prepared uh a demo how to use how to apply uh our tok system to random transactions. And I'm not sure. Oh, I think I cannot play. Thank you. Yeah. So here we provide a simple graphic user interface for to use our uh GB system and for the very first step you need to run the first compiler here and then you need to run setup to commit to the first compiler result. Yeah. And then you just you can just enter your the hash of the transition you want to apply GKP like this. And then you run the second compiler. Then we need to commit to the result of the second compiler. Yeah. And finally you can generate a proof using this pro algorithm about the correctness of the the selected transaction. Yeah. And you can try verify it. Yeah. And you can choose another transaction to run it here. What I want to show you is that the the commitment to the second compilation for different transition can be skipped. Yeah. Because those the second compiler compileration can be reusable for many other different transactions. Yeah. Right. Then uh let me introduce how to how to use our GKP system to for for better Ethereum ecosystem. So uh this is kind of potential application to Ethereum. So just consider consider the situation where we want want to uh run operate some layer 2 channels not the ro of network but just channels and suppose that this channel this channel is uh specialized for limited computations or predefined computations. So in this case we can use our GKP system. Uh we need to run the two-step compilations first right. So the first compilation and commitment can be run on onchain and the once we fix the the type of computation the type of transactions to run to exe execute then we can uh run the second compiler compiler and commit months on the onchain before opening a channel. So once this uh the type of transactions is fixed then we can anyone the users can open a channel their own channel uh on which we can they can run the transactions and generate proofs. Yeah. And in the in the second channel in the in the channel uh the users do not need to uh run any compilers or commitments because they are done in onchain already. And once the users satisfied, users are satisfied with their purpose then they can close they can aggregate the proofs and they they can close the channel uh and they can send back the result to the onchain. Then uh using the just the the single verification equation you can uh check everything about this second the layer two. Yeah, this is this this is our scenario. Right. So here I' like to conclude with some key takeaways. Uh yeah. So when you're designing a GKP or you are unless you uh want to utilize it uh you can't have everything or snacks and stocks have a tradeoff between efficiency and compiler dependency. So uh with our tokashi KP system it is easier to use and faster and lighter uh through our two twostep compilation. So our next move in the future will be to provide a DIY tool uh to launch and that enables any users can launch and operate computation specific state channels or rollups for pre pre to to keep their privacy. Yeah. Yeah. And if you interested you can download a graphic user user interface that we provide and you can look into our go through our paper for theories and also we we have some GitHub course. Yeah, you can find them easily. Thank you. Thank you for your time. Yeah. Okay. Do we have any questions? one question. >> Uh you you said about the proof aggregation on the L2 to L1. How long does it take uh for it to complete? >> Um I'm not sure it's just an MSM. It will be just an M multiscaler multiplication. So it can be done quickly if you use some GPU. >> Yeah. So it depends on the GPU. >> Yeah. Yeah. Right. All right. Yeah. Very quickly. >> Thanks. >> This leads to my question about GPUs and ZK proves being very compute intense. I believe that they're definitely the future, but optimistic roll-ups are quite popular now just because they're easier to implement, cheaper to implement. Um can you just tell us a bit more about this broader um idea of like when are we actually going to have implementable ZK proofs that are comparatively cheap and easy to implement >> just I think we can we can I think we can open some service from next year maybe hopefully. Yeah. So you can meet very easy to use and very yeah you can learn it from next year I think. Yeah. Do we have any more questions? Thank you. Thank you so much. Thank you.