# "Scaling FHE for Global-Scale Private DeFi" by Guy Zyskind // ECC#2 - Buenos Aires 2025

- Channel: [Ethereum Cypherpunk Congress](https://streameth.org/ethereum-cypherpunk-congress)
- Date: 2026-01-09
- Duration: 17:46
- Topics: web3, privacy, now, crypto, cryptography, blockchain, data, security, human right, rights, tech, technology, internet, open source, free, freedom, ethereum, hackers, ethics, cypherpunk, dev, developer, dapp, decentralization, bitcoin, computer, surveillance, cyber, peer2peer, p2p, love, solidity, zk, zero knowledge, education, academy, w3pn, congress, ethereum cypherpunk, buenos aires, argentina, vitalik buterin, privacidad
- Watch: https://streameth.org/watch/yt-zdrvIVYaxEU
- YouTube: https://www.youtube.com/watch?v=zdrvIVYaxEU

## Description

Fhenix founder Guy Zyskind talks about FHE scaling for Private DeFi, encryption, privacy, smart contracts and much more.

Ethereum Cypherpunk Congress by Web3Privacy Now is the world's largest cypherpunk and human rights event.
4500 people gathering in Buenos Aires to celebrate privacy with internet freedom leaders like Richard Stallman, Vitalik Buterin, Roger Dingledine, and Eva Galperin. 

Join us in building a free internet for all.

Website: https://web3privacy.info/
Congress site: https://congress.web3privacy.info/

## Transcript

[applause] Hey everyone. Uh, I know it's late, so I'll try to keep it brief and interesting hopefully, but thank you for staying for this. I'm uh, Guyskin. I'm the founder of Phoenix. We're basically building privacy solutions for Ethereum using uh fully homorphic encryption aka FHE. So I I've been in the privacy space and blockchain for over a decade starting like 2014 and this is the first time I can actually say this confidently that privacy is the new meta and that is a very very exciting uh uh place to be in right now. Now, I'm sure many people here know about privacy and the best way to solve for privacy is via encryption. This is really the cryptography's way to get privacy um since like you know the 20 early 20th century. Now, just to make sure we're on the same page, uh let's do like encryption 101. So, what is encryption? This is the most basic form of encryption. And I'm sure all of you already know this. Um, encryption is really the idea that you can take any piece of data. So let's take this image here. You have a specific key and using that key you can effectively encrypt or mask that image such that it looks like complete random garbage. And unless you have the decryption key um there is no way for you to actually unlock and unmask the data. Now there is a problem here which many of you I'm sure are familiar with. It's that once you decrypt the data it's done. So if you give someone your data and they have the decryption key can they can decrypt it then they can see everything that you do. And so this gave rise to this cool idea called fully homorphic encryption or FHE. This is not a new idea. This idea was actually brought up by Ron Revest. So you might know him from RSA back I believe in the 70s of the last century but this problem was open for like 40 years and only in 2009 exactly when Bitcoin came to be were we able to actually solve it. Now what is FHE what it achieves I mean it's very very hard to pull off but the problem statement is simple. So the problem statement is that what if we had encryption where after we encrypt the data we can actually run any program on top of it and do anything that we want without decrypting. So let's let's give a working example. Let's say we have um let's say we have Alice here. So Alice here has this image again. She wants to encrypt that. That's a private image. She doesn't want anyone to see that, but she wants to modify that image. She wants everyone here to have red shirts instead of whatever they're wearing. So, in today's world with something like Chad Gypt, she can go ahead ask Shaj, hey, take the image and convert everyone's shirts to red. But she does not want OpenAI or the who'sever running the server on the back end to actually see her private data. So with FHE, she could actually encrypt the query, encrypt the image, send it to OpenAI servers, and if they were running FH on the back end, they would have been able to actually modify that picture in place without actually knowing what they're doing. And if it sounds like magic, it is a bit like magic, but FH actually enables us to do so. And then really, you know, Alice is is the only one with the key with the decryption key. She can take the image, decrypt it, and hold and behold, she got the result that she wanted. Now really, FH is ubiquitous like it can do any computation over encrypted data that you want. And if this was free and if this was um scalable everywhere, then everyone would use it, right? Right? Like OpenAI would use it, Google would use it, everyone would use it, but we're not quite there yet. But that's where we should be aiming for. Okay? So, if you saw this for the first time, this should be mind-blowing. If you already knew what FHE is, you should be saying like myself, this is cool. This is definitely something that we need everywhere. But what does it have to do with blockchains? Like why why do we even care about blockchains here? Well, you know, if you like if you like the framing of a blockchain like Ethereum presents, a blockchain is like a world computer. It can run any program and in in a permissionless trustlessness manner when it comes to the integrity of the computation. So everything it does is correct by definition. But on the flip side, and if you're here, you know this, it sucks at privacy. It gives you no guarantees in privacy whatsoever. It's even worse than web two because now everyone can see your data. So really in blockchains we need to figure out a sustainable way to encrypt data and so for privacy and that's where FHE comes in. So let's give a very very simple example. Um here you have a smart contract. Uh who here knows what a smart contract is? I mean just to set the baseline. Okay pretty much everyone. Good. Um, so a smart contract, a program on the blockchain. Here you have an auction contract. It could be an ICO auction to participate in some public sale or it can be an NFT auction. It doesn't matter what you're auctioning. The point is that if the bids themselves are public, then anyone can frontr run it. can see look at the memp pool can front run it and send like these malicious beads that for example grief the user by just sending a a slightly higher uh bid in place. However, in a sealed bid auction where you have all of the bids encrypted, this is no longer an issue. If the bids are encrypted, no one can see it. They really have to bid whatever they believe is the right price. And this is obviously better in many ways, but also this is just like one example. Really, what we believe in is that everything you build on blockchains should have privacy built in. And if I can plug in a bit what we're doing, we are not building an L1, we're not building an L2, we're not building a blockchain. We believe privacy is fundamental and encryption is fundamental at all blockchains at all chains at all levels. So what we're building is coffee which is an FH co-processor that is like an extension to any EVM chain. It can be Ethereum L1 a rollup or even an alt EVM chain. This works with all of them seamlessly. And this is already deployed on on on the Ethereum L1 in testnet and on arbitrum and on base. Now again this is the end of the day. I'm sure you heard of of a lot of applications and things where people want to deploy privacy. You can do things like encrypted ERC20 tokens. You can do things like a private identity and governance and voting. Um, you can even build like privacy preserving AMMs and per and DeFi applications of all kinds. But, uh, if you've heard of FHE before, I'm sure you're telling yourself, well, I heard that FH is slow. And this is something me and my team, we hear every day, every single day by really, really smart people like cryptographers like me. They talked to me in 2025 and are saying FH is too slow. And you can't blame them if I'm if I'm if I'm making a confession. That was me three years ago. So I kept saying we should use tees, we should use MPC. My PhD is in NPC. Like I should I was a big believer in those texts. Um but three years ago I got converted and realized that I needed to update my priors. FHE is actually becoming very fast and it's actually accelerating at a surprising pace and I want to give you an example. So let's do a case study of just how fast FH is becoming. So one thing you need to do a lot with FHE um is is to decrypt. And let me give you an example when you need to decrypt. So this is a wallet, right? This is your plain vanilla metam mask or any wallet. If you're on Ethereum today and you open your wallet, all that happens, this is without encryption. All that happens is you go to some RPC server and you tell it, "Hey, I want to read my balances from the database and that's it. It's done. No problem." But what happens if all if all [clears throat] of those balances are encrypted with FHE? Now someone needs to decrypt those entries for you before you can see them and they need to do it in a way such that no one not even the validators in the network not even the RPC node no one in the supply chain except you the user learns the decrypted result. So without getting into too many details, the way you do that is via a cryptographic protocol called threshold FH decryption and that protocol is slow or it was slow until literally just a year ago. So if we were standing here last year talking about that then I would have to concede and tell you that for a single decryption you're going to have to wait about half a second to a second and you would be able to do at most 10 decryptions a second. Now let's play let's play a game. Let's see if this is even workable. So imagine you have a thousand users using FHA using this system. Each one of them has 10 uh 10 assets 10 um 10 tokens encrypted that's already if they're coming into the system at the same time that's already 10,000 decryption requests. And with those kind of like with that kind of throughput, users would have to wait like 30 minutes and even more just to get a response. So this is obviously unusable in DeFi. Like you can't use that on Ethereum. You can't use it in DeFi. Even today it's I would argue useless. So what we did is we said okay this is unworkable. We need to unpack this and we need to solve this. And what we did was we went through the the the the literature around threshold FH decryption. There's 15 years of literature and all of them used a single approach. That approach was that it's called noise flooding. Noise flooding again I don't want to get into too many technical details but if you allow me just briefly. So in FHE every cipher text the way it's protected or one of the ways it's protected is by adding noise to it. But that noise is sensitive. If you ever get if ever someone learns the noise of the cipher text, then from that they can mount an attack and actually learn the private key, the private decryption key, which is which means that like security and privacy is completely broken. So what everyone until us did with noise flooding, they essentially they blew up the cipher texts. They made they took the cipher text they made them much bigger and in computers when you operate on much much bigger words or much bigger like data structures everything becomes much slower. So that was the bottleneck and so we said okay let's take an MPC approach to this instead um and let's not use noise flooding at all and let's just make the MPC protocol to clean out the noise very very efficiently and when we did that we got like amazing results so you know this is 2024 a year ago what I told you this is today so like you can actually do like less than with the right network less than 10 millisecond latency per decryption. But this is the important piece like you can do tens of thousands of decryptions a second and this can already like this this gets to like visa level scale if you will. Now um if you want to see the technical details if like you're more technical or cryptographer there's [clears throat] a paper out uh in e-rint you have the link here. Uh this was this is very new. This was presented a month ago in CCS, one of the top academic uh cryptography conferences in the field and it actually won uh one of the best paper awards. Um and this is this is not just theory. This is already implemented in our stack. So if like you're building with confidential D5 primitives, it's already powered by this. Now I do want to move on from technical a bit higher up and and talk a little bit about what does that mean in practice and really what that means in practice and one of the nice things about FHE is that it's like very very easily programmable and you can actually get the level of um user experience that you can with even even if you're not using privacy and this is a big pain point I know that anyone here who use rail gun, tornado cache before, zcash, all of those, you know how painful that user experience is. And nice thing about MPC is that you can really really uh uh do much better. I'm uh can we open this link? Can we open the link? Yeah. So this is an example of just how sleek the UIX is. Like it's really really impressive. All of this is running live. This is non mockup. All of everything is encrypted. Everything is running FH under the hood. So the the one before was NFC. This is this is using Bluetooth to transfer encrypted tokens. Okay, so two take-h home messages here. FHE is not slow. It's already usable today. You get amazing UIX. If you're a developer, you can play with this today. We're not doing a chain. We're not trying to sell you like using something else. This is about adding this to Ethereum and blockchains in general. And on Wednesday, we're doing encryption day, which is going to have not just us, but like many of like the top figures in the industry, talking about all the best and latest in like ZK, FHE, MPC and all that. And then I will also talk a bit more in depth about like what's next in the next year and how we are making FH scalable not just for blockchains today, but like for Visa scale payment networks level. So thank you very much and you know if you want to talk about this I'm here and and the entire week and our team is here. So thank you [applause]