DSS 2026 — Are MPC wallets ready for the post quantum age? Speaker credit under review Source video: https://www.youtube.com/watch?v=ndaxy2I5gtE Automatic captions — not human-reviewed. Complete spoken-audio coverage is unverified. Source language tag: en (English). Source export SHA256: 54063e7345327fdb7383b24447ffe5503744dc387fa6079dfff1577cbb7189db [0:00] Per so good afternoon everyone. I'm Yun [0:02] from Fortify and today we'll be talking [0:04] about whether your MPC wallets are [0:06] actually ready for the postcoh. So to [0:09] starters we are fortify by paxos. We do [0:12] NPC wallets that for in institutional [0:16] usages where NPC is at our core that [0:19] then keys are split and never [0:21] reconstructed to sign. Uh disclaimer [0:23] this talk was meant to be a a joint talk [0:26] with my CEO Dimma but he wasn't able to [0:28] attend. So today we will discuss why the [0:31] heck should we care about quantum [0:32] computing? What's the actual quantum [0:34] thread and what are the new signatures [0:36] that we're considering? what the [0:38] blockchain are planning to do and how [0:40] does postquantum look in the NPC world [0:42] and well if it's not ready what can we [0:44] actually do cool so why the heck should [0:47] I care now to be honest it's quite [0:49] simple quantum attacks are getting quite [0:51] closer now like large companies such as [0:54] Google are targeting 2029 to migrate to [0:56] fully postquantum schemes and ECDA is [0:59] going to be dupregated by 2023 by 230 [1:02] sorry so it's quite around us around the [1:06] corner to be honest And moreover, [1:08] research and standardization of new [1:10] schemes actually take time, especially [1:12] the MPC schemes that really comely come [1:14] after only the signing schemes. And [1:17] moreover, the chain migrations also take [1:19] time. So, we need to start early. But [1:21] what actually happens once we have a [1:23] quantum computer? First and foremost, [1:25] almost all public key cryptography is [1:27] going to be broken. Using Shor's [1:29] algorithm, we given a public key, you [1:31] could simply take a look at it and say, [1:33] well, I can break that and find the [1:34] private key quite instantly. So that [1:37] means that all our ECDSA, EDSA and other [1:40] elliptic curves or even RSA is going to [1:42] be broken. And on the other hand, [1:44] there's also the symmetric keys such as [1:46] SHA and AES. They're going to be [1:48] weakened, [1:49] which means that we need larger keys, [1:51] but we can still use them. So the major [1:54] blockchains are using skins that are [1:56] going to be broken. Then we need to [1:57] replace them. But what are we going to [1:59] replace them with? Well, there are many [2:01] options er such as Magnus actually said [2:05] there are many schemes such as hashbased [2:07] signature which we can use specifically [2:09] SLHDSA and there are the latest ones [2:11] such as MLDDSA and Falcon which these [2:14] schemes are the most [2:16] used schemes to be honest the most [2:18] looked at because they are the N [2:19] standards there are other schemes which [2:21] we won't discuss and there are other [2:23] schemes which were really leading in the [2:26] candidates of the N standardization and [2:27] they were simply broking along the Okay, [2:30] so you can think of it as it's quite [2:32] hard to actually think of and invent new [2:34] schemes and new cryptographic [2:35] mathematics. So I won't discuss the [2:38] hashbased signature a lot again thanks [2:40] to Magnus which helped me a lot here [2:41] this time but in very short terms you [2:44] have a hash value you have a secret key [2:46] you simply hash it a number of times say [2:48] hey this is my signature and then you [2:50] give it to some verifier which hashes [2:51] the remaining number of times compare it [2:54] to your public and says well you're okay [2:55] that signature is fine on top of that [2:58] you build more sophisticated scheme such [3:00] as SLHDA which I I like to think of it [3:03] as a tree signing the root of other [3:05] three. Now a very important point here [3:08] is that it's considered a conservative [3:10] option. Why? Well, we are relying on [3:13] hash hash schemes which we know and love [3:15] and use every day. And moreover, as you [3:18] can notice, I never said which hash [3:19] function we actually are using. So if [3:21] the hash is broken, simply change the [3:23] hash function. But you don't need to [3:24] change the scheme. Now it's not all [3:27] perfect. It has its own weaknesses such [3:29] as being slow and also it has large [3:32] signatures. like it's almost 100 times [3:34] larger than SDSA. [3:37] On the other hand, we have the latises [3:39] which is the new cool method on the [3:41] field of cryptography. What's lises? [3:43] Very complicated. In short bases, it's [3:46] simply think of it as linear algebra. [3:48] You have your basis vector. It's long [3:50] and skewed and angly. And then your goal [3:52] is to find a very small vector that [3:54] simply is hard to find giving those [3:56] bases. In the picture here, we see that [3:58] it's 2D. Actually, it's a thousand [4:00] dimension. So, it's quite hard. And [4:02] there are the most two interested [4:05] schemes which are MLDDSA and Falcon. [4:07] LDSA relies on noisy linear equations [4:09] and Falcon relies on structured latises. [4:12] Where are these? Well, that's not [4:13] important for our talk today. And [4:15] there's a very interesting here point [4:17] here which is these math are actually [4:19] quite debated. Like if you ask the [4:21] German federal office, they will say [4:23] well it's nice, it's cool, but please [4:25] use it in a hybrid form. Please use some [4:27] kind of ECD say currently because we are [4:29] unsure that it's actually safe. or even [4:31] there are people in the industry that [4:33] simply say well please avoid structured [4:35] mathematical assumptions entirely. I [4:37] think it says something a bit or two [4:39] about these schemes and also it has its [4:42] own weaknesses such as being large in [4:44] public keys but it also has new [4:46] strengths which is being fast and being [4:48] small in signature size. Cool. Now let's [4:51] take a step back and look at the point [4:52] of view of a blockchain. Well, what's [4:54] important for them? There are three main [4:56] points. First is the hardness which I [4:58] find the most interesting as a [4:59] cryptographer. And there is also [5:01] signature size which and key size also [5:03] which is quite important for them [5:05] considering that we want small [5:06] signatures and keys. And there's also [5:08] the aggregation points. So what each [5:11] blockchain is considering? Well, it's [5:13] quite diverse to be honest. They are the [5:15] ones that are considering hashbased [5:16] because of its conservativeness such as [5:18] arc essentially which I really like [5:20] because you can now use pre-ompiles of [5:22] SLHDSA specifically and their mainet [5:25] which is quite nice and there's bitcoins [5:27] which in simply are considering a forked [5:30] of SLHDSA [5:31] similar to sphinx minus essentially [5:35] which reduces the size signature sizes [5:38] and also make it faster. On the other [5:40] hand, there are ones which are [5:41] considering the fingering, sorry, [5:44] favoring the size such as Salana and [5:46] Tron and Ethereum. As Magnus again said, [5:49] they are quite generic. They're saying, [5:51] well, use what the heck you ever you [5:52] want. You can use Falcon, MLDDSA, ECDA, [5:55] sorry, SLHDSA. Eventually, currently, [5:58] you can actually use it in Ethereum, but [6:00] eventually it will be native with [6:02] pre-ompiles with EIP8141. [6:04] And for the consensus, they are [6:06] considering mostly the hashbased again [6:08] for its conservativeness. [6:10] Now let's even take a step back. What [6:12] are what is going to happen with NPC [6:14] wallets? Well, a bit about NPC wallets. [6:16] For those who don't know, NPC is simply [6:18] the idea of having multiple parties [6:20] perform a computation where your inputs [6:23] are simply private. Think of it as a [6:26] multiple parties which know which want [6:27] to know who is richer to be honest like [6:29] it can be any math problems that you [6:31] want to solve. And in specifically in [6:33] the signing world of crypto, it's the [6:36] idea of having number of parties which [6:37] each hold a share of the key of the [6:40] private key and eventually they need to [6:42] communicate together perform some kind [6:43] of protocol computation eventually to [6:46] create a new standardized a new standard [6:48] signature which is invincible on chain [6:50] compared to some something like multisig [6:52] which you know how much people actually [6:54] were used to sign that signature. And [6:56] the primary bottleneck here is actually [6:58] the communication. The math here, how [7:00] much it takes to create a signature or [7:03] to do the computation relies heavily on [7:05] how many rounds do I need to send to the [7:07] other people. How many packets do I need [7:09] to send them? Because you're doing a [7:10] joint computation. You don't simply send [7:12] a message and then boom, the secondary [7:14] is ready. So there are three main groups [7:17] of computation of steps essentially. [7:19] There are the linear steps which are [7:20] simply local operations that you do. [7:23] They're quite cheap. They're local [7:24] operations. you don't need to send [7:25] anything to the network. On the other [7:27] hand, there are some things like multip [7:29] multiplying secrets. These require [7:31] simply a round or two based on the [7:32] scheme of MPC that you're using. And [7:34] then there's the interesting ones, the [7:36] nonlinear operations such as hashings [7:38] which require a lot of communication [7:40] rounds to do. Essentially, you would [7:42] need a lot of communication rounds [7:43] sending to each other waiting for the [7:45] response until to be able to perform [7:47] that signing. And the important point [7:49] here is today's signatures they are [7:51] using mostly linear operations. So they [7:53] are quite interactive to be honest. But [7:55] the postquantum schemes well there are a [7:57] different story. They are more nonlinear [8:00] rounds and they require a lot of more [8:03] nonlinear operations to do so. So you [8:06] can assume that's going to be quite [8:08] slow. How slow? Well, that's exactly [8:10] what we wanted to see. And we decided to [8:13] take a look of what would happen if now [8:15] there was a quantum computer. What how [8:17] how would we actually do and perform [8:20] with these? So we took the most [8:22] important known protocols in the MPC [8:24] world considering the postquantum war [8:27] and benchmarked them benchmarked them on [8:29] the two-party signing. We considered two [8:31] different deployment models. First the [8:33] same region which we have two servers [8:35] sitting next to each other which high [8:37] with high bandwidth network and on the [8:39] other hand you have the phone and cloud [8:40] similar to how we do at Fortify which we [8:42] have an enclave and a mobile which [8:43] communicate and here your bandwidth is [8:46] quite more constrained and we time two [8:49] things. First the offline the offline [8:51] phase which is simply preparation of [8:53] your message that you're going to sign. [8:56] You can do this before you need you know [8:57] what message you're going to sign which [8:59] in case of blockchain what is your [9:01] transaction that you're going to sign [9:02] and the only phase that once you have [9:04] your transaction that you want to sign [9:05] how much time would you need until you [9:07] sign it. So now we're going to go over [9:10] the most important protocols and see how [9:13] it actually deployed. So starting with [9:14] MLDDSA there is the myth world protocol. [9:17] Quite a nice one. And the problem with [9:19] MLDDSA is the idea that even once you [9:22] have the signature of your transaction, [9:24] you can't simply broadcast it away [9:26] immediately. You need to check that it's [9:29] it has some kind of requirements that [9:30] doesn't leak your information. And the [9:33] idea here is that in the NPC state, in [9:36] the NPC regime, you need to actually [9:37] verify that it's okay to rebroadcast, [9:40] but you have other people that you need [9:42] to also trust. So you can't trust them. [9:44] So what did they do? They essentially [9:45] took a different approach was simply [9:48] each party would consider his own part [9:50] of the signature and say okay it [9:52] verifies that it's small enough let's [9:54] say and then broadcastes to the other [9:56] people to simply build up the signature [9:59] which enables it actually to be quite [10:01] interactive on each and every different [10:03] network and eventually it's quite [10:05] practical up to the six parties but you [10:08] know most of the deployment modes is [10:09] actually enough for and on the other [10:12] hand there is threshold falcon threshold [10:14] falcon is quite write new protocol and [10:16] it's a breakthrough to be honest for [10:19] those who know about about Falcon. Well, [10:21] Fal Falcon relies on floating points. [10:23] Floating points and NBC are they aren't [10:26] friends. They hate each other. They [10:27] don't work together at all. So, what did [10:29] they do? They essentially targeted a [10:31] different random sampler. Instead of [10:34] Gaussian, they used Klein. What does [10:35] that mean? Not that important, but it's [10:37] simply that they use integers and they [10:40] also target batch signs. The number that [10:42] we are seeing here are essentially when [10:43] you want to sign a single message and [10:45] you can see that it's not quite so [10:47] interactive such as MLDDSA but still [10:50] quite usable on the server server side [10:52] on the mobile and server well not so [10:55] much but if you consider bat signing [10:58] like let's say I don't know signing 200 [11:00] signatures at once it can be actually [11:02] quite usable mostly on the server server [11:04] side and there's a small caveat which [11:07] you need a trusted dealer like some kind [11:09] of third party which comes and splits [11:10] your keys and shares and does all the [11:12] first pre-processing state. And now the [11:15] fun part begins for me which is SLHDSA. [11:19] Up until last two months, there weren't [11:22] any MPC protocols at all for SLHDSA and [11:26] there was actually a paper which [11:28] released which used fully homorphic for [11:31] for those who know which required eight [11:33] GPUs such as RTX 490 and could sign a [11:37] message in 3 minutes. Well, nice, cool, [11:40] but not so interactive. And we have been [11:42] doing doing our own research to be [11:45] honest and as a HD and we were able to [11:47] sign in two party mode between a server [11:49] and a server or a mobile and a server in [11:51] less than a second. And how did we do [11:55] that? Well, we're going to publish a [11:57] paper later on in a couple of few weeks. [11:59] So stay tuned for that. But in short, [12:02] some of the key ideas that we did is the [12:04] fact that SLDA can be quite paralyzable. [12:07] So what happens if you paralyze it with [12:09] GPUs and quite heavy optimizations? [12:13] Well, you can get some interesting [12:14] things. What are the caveats from our [12:16] side? Well, we're mostly targeting the [12:18] two-party signing although it can be [12:20] done also for multiple party and we're [12:22] targeting the shape parameter sets for [12:24] those who know that hash function and [12:27] specifically because it's quite MPC [12:29] friendly. [12:31] Now let's say that today a a quantum [12:34] computer is deployed and a lot of the [12:36] other schemes other than the ones that [12:38] we log they don't have an APC protocol. [12:40] Well can what can you do actually [12:41] instead there are many options such as [12:43] onchain multiig but then you would make [12:46] it visible how many signage you need and [12:48] how many verifiers you need or secret [12:50] sharing. What if you simply take your [12:52] private key and secret share between [12:53] number of parties and simply have some [12:55] kind of trust in environment signing [12:57] that it's okay but you know if your [12:59] trusted environment is compromised then [13:00] your key is compromised not as NPC which [13:03] you need to compromise all the parties [13:05] at once there are also snack [13:06] aggregations nice but they're quite [13:08] heavy and require specific operations on [13:11] chains and there are hybrids which I [13:13] really like I even like it so much that [13:16] we're going to look at a case study what [13:18] happens if we actually need to use [13:19] hybrids we can do something quite [13:21] interesting. You can use the MPC for the [13:23] classical side for the ACDSA side and [13:26] then you can also use secret sharing [13:28] specifically to pro like to secret share [13:31] your postquantum key and then use it on [13:33] a trusted environment. Then you would [13:35] actually need like a way to break let's [13:38] you still have the NPC because you need [13:40] the classical part to sign with to sign [13:42] the classical part like you need the NPC [13:44] on the other side for the postquantum [13:46] you need the two secret shares. Now [13:48] there's the point that it's [13:49] reconstructed on the trusted [13:50] environment. So if the trusted [13:52] environment is compromised, you still [13:54] need the hybrid signing unless a quantum [13:57] machine actually compromises you, then [13:59] you're simply screwed and you can't do [14:01] anything. So these are the points that [14:03] if we actually consider [14:06] more schemes essentially require more [14:08] NPC research and the alternatives are [14:11] not exactly a replacement. Each one has [14:13] its own problems and chains should [14:15] actually weigh NPC when they choosing [14:18] the scheme because when they're choosing [14:19] the schemes they are quite determined [14:22] how much the MPC wallet can actually [14:24] follow in the postquantum world. Yeah. [14:26] Thank you.