Wave 1 happened in 41 minutes. Not one of those UTXOs was created before March 17th, 2021, when the Coldcard firmware bug was introduced. This is wave three, much more sophisticated. I have direct victim confirmation for about 400 of the 1,500 or so that we currently call high value. The prudent measure for any sort of financial institution when receiving a credible cryptographic forensic report: freeze it on your platform.
Alex, the vibe has shifted from the last time we saw each other, which was a Thursday at the Galaxy event in DC.
Yeah, it was just shifting. I think you came maybe an hour before that Pub Key event started and said there might be some issue with Coldcard.
And in my head I was like, you know, oh no. But I was hosting that event, so I got all tied up and I didn't realize that I didn't really see you the rest of the night. And then the next day I listened to your episode with James OB where you said that you had to be in a corner on the phone the whole time, which I of course now totally understand.
It's been a shitty week. But I wanted to get you on because you and the team at Galaxy have done an incredible job of tracking the attackers. It looks like there's multiple waves of attacks of people brute forcing private keys of those that were generated on Coldcard devices and contacting victims. And I think the knowledge you guys have gathered and the information that you've gathered over the last five days is important for people to understand what's going on. And so I have the chart here. Maybe we start there. Just what you guys have found based off of the information and the outreach they've gotten from victims. And I think you said so far 94 have reached out to you.
Yeah, it's in the mid 90s. I mean, it's growing all the time. I mean, especially as I did Nick Batia's show yesterday, I did Unchained yesterday where of course I called for victims to not be ashamed. You did nothing wrong. You did not deserve this. First of all, file a police report. Well, first of all, secure. If you haven't secured your funds on a potentially vulnerable Coldcard, please do that immediately. But also share your drained addresses and the transaction IDs that drain them. I know it's hard for people to do it, but because I've been saying that, people have been sharing, and it's because of that sharing that we've been able to identify so many of the coins and where they are, what addresses they're sitting in. So, you know, it's also sort of in exchange, although I don't think of it that way. In fact, initially I wasn't even doing it this way, but I can fulsomely trace anything in Bitcoin. I have a very powerful proprietary Bitcoin stack with some clankers sitting on top of it that help me traverse. So I've been providing back forensic reports to victims that they can then use to report to their local authorities, to the FBI, and that we can use as well to report to crypto exchanges and chain analysis and all of that so that there's a chance that the funds get frozen. But please do continue to reach out and share those addresses with me so that we can keep building out the picture of the attack as it unfolds.
And let's dive into the nature of these attacks. Obviously, it started late Thursday afternoon, early Thursday night, and I think it's become clear that AI was used to identify this vulnerability and then exploit it. And you can sort of see that in the patterns of the waves of drainings that we've seen. In the first wave, it seems apparent that the person who started draining wallets of Coldcard users didn't really understand best practices or how to actually scope through a wallet. There was a very specific pattern to this first wave. Correct?
Yeah. So this pattern, I don't think they were the first. There certainly where I learned of the pattern was the engineers at Block Inc. Obviously the guys behind Cash App, Bitkey, Spiral, etc. Square, right? They identified this pattern as a pattern, right? A key piece of the pattern was a fixed 30 sat per vbyte fee on all draining transactions. That was substantially higher than the median fee rate at the time, which being higher than the median is not in itself that surprising, right? Attackers are willing to overspend to make sure they get the funds. But normal people with urgent transactions don't have many bursts of transactions from previously unknown addresses with a fixed fee rate, right? Like your typical wallets will suggest fee rates, you know, if you want it confirmed in this amount of time, right? Fee estimation is actually a big thing Bitcoin Core itself has worked on, let alone all the other wallets, right? So that was a key feature of the pattern of wave 1. This is the only pattern of coins we have very high confidence in are stolen due to this vulnerability that was solely based on a pattern, right? And so they identified this pattern. It was very mechanical looking, right? There's a lot about it that looks like it was designed and then executed by automation, I'll say, right? And there are other patterns that also look like that, right? They don't all look like that, but many of them have features in the topography of the movements and the design of the transactions themselves, the fee rates that look automated.
What in terms of where these coins are going to? I think the first wave had a lot of bad addresses too.
Yeah. So waves one and two, and in fact there's almost a better chart that I tweeted earlier. Well, I have a chart of wave one as a perfect example. Wave one alone, these are like called Sankey charts, and they show fund movements. This is like just the highest level overview chart, right? This shows the various waves. Waves one, two, and three are high confidence promoted. That's where people are getting this like 13, you know, 56 type number. I've also promoted a lot of what we call footprints A through N. Some of them are pattern matched. All but what I should say is that wave one was designed and then later confirmed, but first designed solely by pattern matching from Block. I took the pattern they described they saw, I set off across the entirety of not just confirmed blocks and their transactions but also the entire UTXO history. So that's not the UTXO set. That's the UTXO set at every single state in Bitcoin's ledger ever, right? It's literally like 3.8 billion rows that I have. Yeah. So if you look at wave one here, this is all wave one now. Like each of these four funnels, right? These are many addresses in each case being consolidated into four independent what we call collectors, right? They're collectors because they consolidate stolen funds into one step, and then that collector sent to three second hop holding addresses. And you can see for some reason there's this little gray band that comes out of the first funnel here. That's not another second hop. That's actually still sitting in the collector, but I put it on the right side here just to show that like the right side is where the funds sit. So this is actually a pretty typical looking siphoning topology or topography. I'm not sure which. I think people know what I mean. That you see in like other types of crypto hacks, right? It all gets drained immediately into an address the hacker controls, and then often they move it to like the proper setup that they want, right? Now sometimes what you see, especially in a hacker determined to exfiltrate the funds into the fiat system or into whatever other chosen currency they want, then you start to see radical things like peel chains where they blow up the held coins into like hundreds or thousands of different pieces, and then later they rejigger them into 20 pieces that don't look the same. They do it again and again, right? Just to make it really difficult to track. Also often they siphon them into like if stablecoins are ever stolen, they often immediately transfer them into ETH, right? Because it's more immutable than the stablecoin. Or bitcoins sometimes immediately get sent through Thorchain and then sent onto ETH and then sent through another bridge into this thing because bridges are harder to track funds through. Unlike those, all of the coins here in wave one are just sitting inert in these three addresses on the right. So they have not moved, right? And this is also true for waves two and three. But one of the other things that's interesting here about wave one, and I didn't in the thread where you pulled this, don't have the wave two. Wave two looks quite similar as well, and where it's many victim addresses consolidated into a holding address and then moved to a second address where they currently sit inert. That is a very clear pattern, right? Like wave three, which is the other one that's in that thread, just maybe if you pull up as an example, looks totally different. And I will say also, there is no co-spend between waves one, two, or three attacker addresses. There were co-spends between those bottom two funnels. So this is wave three. This is actually much more sophisticated. There is no common collector address. This is 293 chains, and by chains here we mean like inputs, groups of inputs. So 293 transactions funding 293 staging addresses funding 293 P2WSH vaults, right? So each group of victim addresses was by itself moved into a staging one, and then by itself moved into a vault which, you know, could be a multisig. We wouldn't know until they spend transactions, but none of those have been spent. I think all but six of the 293 staging addresses only contain funds from one wallet, and now we can see that because of like co-spends of the wallets, and I also have a lot of victims that I've identified that are in these. Right, this is how we... And so the other thing is like wave one was a pattern that was described by Block, which I expanded, and later has been confirmed by many victim testimonies, reports. Waves two and three were identified by Galaxy Research solely by victim reports, right? Like people started saying my coins were lost. And when I gathered like 10 from wave two or like 10 from wave three, it started to become apparent. It didn't actually take 10. And it takes fewer than you realize because once you get like two or three that look alike, you send the clankers off to look at all the blocks all around that and say, 'Does anything else look like a bunch of addresses into a staging address into a, you know, pay-to-witness script hash, whatever, into a vault?' Right? Then you say, 'Oh my gosh, there's an entire burst of transactions that look exactly like that.' And that's how you know. And then of course I start publishing about it. People start saying, 'I think I might be in wave three.' They send me their stuff, we get further corroboration, it becomes like high confidence. So, you know, one another interesting feature of one and two, I told you they look similar, right? Many victims into very few staging addresses, then into second hops where they're inert. The only two of one, like of really there's a couple down, talk about footprints, which are these smaller operators that have emerged, but there are a pretty sizable number of people who were drained in wave one that were not completely drained who were later fully drained in wave two. So there's a number of those that gives us some, I would say, medium to high confidence that wave one and wave two are the same attacker. Now, you know, someone could, and by the way, wave one, to your point, was all just after midnight July 30th UTC. So some people were like, 'Oh, I was July 29th.' And it was like, 'No, if you look, you were like, you know, West Coast, July 29th at night. That was actually early morning UTC July 30th.' We use UTC because that's like the default in Bitcoin Core and stuff. So it's just like easier to, you know, pick one. Wave one happened in 41 minutes. Very identifiable, right? Like you don't usually, because another big identifying feature is that not one of the high confidence, what I call promoted, meaning like we're saying these are drained, like this is no longer a question, not one of those UTXOs was created before March 17th, 2021, when the, you know, Coldcard firmware bug was introduced. And I will say like there's other patterns, like the median dormancy of all these coins is like four years. An enormous portion of the coins in waves one, two, and three, which are the majority still of, you know, high confidence victim addresses, the vast majority had never spent a coin, right? So these are just receiving coins, very indicative of people stacking sats into their Coldcard and, you know, in cold storage. So, you know, I think like there's co-spend which can help identify, but again, everything after wave one were pattern to the extent we've found patterns. There are some where I'm not seeing a pattern, but I have a victim report it, the one that I see it looks credible and drained, and so we've incrementally added those too. But also we found something like we're up to, I think in the graphic, the first graphic you showed, we're up to footprint N. That means we have multiple, the footprints are we have multiple victims where the topography looks similar, they don't necessarily co-spend, a few of these co-spend, and so we're very confident and have promoted. But as you can see, this graphic is everything that we haven't yet confirmed. This goes up well over 2,000 Bitcoin. The shaded ones are I think not confirmed, right, by any owner, right? These are patterns that we found but haven't yet gotten a confirmation at all. Some of those footprints are, they're identified by the fact that victims showed us and we've, you know, said, is there anything else that looks like this? But we don't quite have enough confirmation that we should extrapolate it out to transactions that we don't actually have confirmed by a victim, right? Because there are plenty of transactions in waves one and two and three that we haven't actually gotten a victim report about, but we're very confident are part of the wave. So that's just some background on the methodology that we've been doing. And just as background, like I have direct victim confirmation for about 400 Bitcoin of the 1,500 or so that we currently call high, you know, value promoted. So a sizable amount I've talked to.
Yeah, it's so heart-wrenching.
But and I think explaining, if you haven't, if you have a Coldcard, I don't care what your setup is. I mean, if you roll dice, you should be fine, but if you don't have confidence that you did it correctly, just get it off.
Yeah, and one another important thing to say, but again, and I have a Coldcard MK3, I believe, with no passphrase and no dice in a multisig quorum that cannot reach signing threshold. So it really can't be affected at the moment. And in that quorum I've never spent. So there the address isn't known to hold coins on chain, for example, because it's, you know, like a script hash address that has never revealed its script, right? So, but I haven't, because I know that that isn't vulnerable at the moment, I haven't actually rotated the Coldcard out yet, but I absolutely will.
So, and I will say, I haven't updated this analysis since like midday yesterday, but I can say in waves one, two, and three for sure there are zero multisigs identifiable. So as far as I'm concerned, there's no evidence that any multisig setup has been breached here. Now, of course, if you had like say a two-of-three multisig quorum where two of three are Coldcard MK3s like mine with no passphrase and no dice roll, that is vulnerable. I haven't seen one instance yet of it being attacked. It's probably, you know, Rob Hamilton and the red team working on this, and James OB, are probably more apt to actually talk about that, but I understand that's probably lower on the tier of priorities that the black hats are looking at to search namespace for. It's more complicated. But all of that being said, like I am of the view that like anything on a Coldcard at all, including my sub-threshold quorum key, should be rotated. I mean, there's just no reason to stick around if you have any doubt whatsoever, in my opinion.
Sup, freaks. This was brought to you by our good friends at Square. If you run a business, you need payments, you need hardware, you need software, invoices, point of sale tools, and a system that does not turn every basic operational task into a headache. Square spent years making it easier for small businesses to get paid and keep moving. And now with Block leaning deeper into Bitcoin, Square sits at an important intersection: real world merchants, payment infrastructure, and the future of Bitcoin commerce. We're making Bitcoin everyday money, freaks. If you are starting or upgrading your business setup, you can get up to $200 off Square hardware at square.com/go/tftc.
All right, freaks. You know, I don't take sponsor money from products I wouldn't use myself, so listen up. The Aaven Bitcoin Visa card is one of the most interesting things I've seen in the Bitcoin lending space in a long time. Here's the deal. You can get a line of credit up to a million dollars backed by your Bitcoin without selling a single sat. No gains, no annual fees, no minimum draws, and your Bitcoin is custody by BitGo, one of the most trusted names in digital asset security. Aaven never lends it out. There's no rehypothecation. You stay in control. You can lock in a fixed rate for up to 10 years. 10 years, that's 10 times longer than most lenders out there, or go interest only for up to five years. Rates start at 8.99% APR for a product that lets you keep your stack and still access liquidity. It's hard to beat. On top of this, guess what? You also get 2% unlimited cash back every time you use the card. Spend fiat, keep your Bitcoin. That's the whole game. If you've been stacking for years and you need liquidity without triggering a taxable event, this is worth a serious look. Go to aaven.com/bitcoin. That's aaven.com/bitcoin. Check it out.
Yeah, and it's a timing thing that I mean bringing this back to the waves. Obviously waves one and two had a very specific pattern, but that is the game theory at play right now. It's just a ticking time bomb. If you have private keys generated with the affected versions of the firmware, which started being released in March of '21, so I think version 4.0.0 and beyond are affected. Different models have different levels of entropy. Older models, I think, had more, I'm pretty confident, have more entropy, but still not a sufficient amount of entropy to protect you from a brute force attack. So move your coins if you haven't already. And the game theory is such that once the alarm bell was sounded last Thursday and became obvious that this was exploitable, you just have to assume that other actors are going to enter the fray to try to exploit this too. So the time bomb could be accelerating. The pace of the clicking, ticking of the clock can be accelerating. So move with haste. And I think that's another weird factor that's entered the conversation too is there's, I mean I've been listening to Spaces and following this discussion since Thursday night on the train ride back. People are already talking about this because you had people like Wicked Bitcoin, Portland Hodl, Rob Hamilton, Praine, like basically downloading the script to brute force private keys or seed phrases on their computers, and they were able to identify wallets. And the ethical conundrum of do we as white hackers sweep into addresses that we control in the hope that we can get this back to the original owner. Conversation started pretty early Thursday night, and I think there has been confirmation that some people have engaged in white hat sweeping of these coins, but it gets into again the ethical dilemma, but then just the logistics of even if you do that, how do you get the coins back? And I think that's another important detail. In if there is a scenario in which we identify the attacker or white hats have swept funds and they want to return them, I think the ability to prove that you were the individual that actually created the private key initially depends on you actually having your physical device.