Large-scale Coldcard compromise (1128.47 BTC stolen so far)

[oll]

In my locale, we always come to the conclusion in discussions that open source wallets are better than closed ones, obviously because they do not conceal surprises. But in the age of AI development in open source, if people don't see mistakes, AI will see them, which we witnessed (Coinkite wrote about the use of AI by intruders). If this vulnerability has not been found in the five years of its existence, it has given a false feeling that the code is perfect..
 Of course, the developers are also to blame for making the low-entropy decision. But I think that all developers with open source code should not disdain unnecessary code reviews, both by their own specialists and by the most intelligent AI...

[DireWolfM14]

I just want to know from DireWolfM14, how does it develop if you are still using this wallet?
I hope your BTC is not affected by this case.

I do use my ColdCard as my primary hardware wallet, but none of my seeds were generated on it.  I used a different method/device to generate my seeds and I use strong passphrases on every wallet, including the ones that are mostly transitory.  When generating the seed for my cold wallet I added analogue entropy as well.  So far so good, my bitcoin is still sitting pretty, AlhamduliLah.

Definitely a scary situation, though.  I feel bad for those who lost their stash and hopefully the hacker gets busted.

[greysonz]

It's very concerning, I see Coldcard wallet is recommended by Jameson Lopp on his blog too.
https://www.lopp.net/bitcoin-information/recommended-wallets.html

~

Will it be time for him to remove Coldcard from his Recommended wallets list?

Ledger should also be removed because of that seed phrase recovery update they introduced some time ago.

Now the phrase in the phrase "your keys - your coins", the first "your" now means only self-generated, by own hands

[goldphysicalbitcoin]

The actions of the malicious attacker are truly regrettable. For those new to Bitcoin investment, purchasing an ETF or using a custodial service like Coinbase is likely the most suitable option at present; before the advent of mainstream hardware wallets, self-custody relied primarily on paper wallets—a method that remains secure to this day—whereas hardware wallets introduce additional risks that cannot be overlooked.

[PrivacyG]

This particular bug comes from a weakened entropy source. They used some Python-based RNG instead of their on-device more secure hardware RNG by accident while replacing the GPLv3 code (their obsession with GPL is ultimately what sunk them in the end).

You would think a company that produces Hardware Wallets and develops their Firmware would audit the code before publishing a new version for everyone else to see.  I am not an expert coder my self but looking at the way this vulnerability is described, it sounds like a really stupid mistake that would be easily recognized by an advanced developer like those I assume are developing at Coldcard.

-----

That just doesn't make sense! Anyone with the skills to pull this off shouldn't make such a rookie mistake, unless the account is owned by a scapegoat.
Or could it just be an amateur who's in way over his head? Someone who accidentally stumbled upon this weakness, and suddenly ended up with $70 million in stolen funds?

Is my mind playing really bad games on me right now or years ago there actually was a couple that was showing off their suspiciously luxurious life style on the Internet and they ended up being exposed as the actual thieves of one of the largest Bitcoin hacks in history?

There are so many options!  Such as planning to leave the country with all that Bitcoin to some where you can not be extradited from and being caught right before finishing the plan.  To be honest with you, having the intelligence to steal in a way others did not find yet does not also imply having the intelligence to not get caught.  In fact.  Thieves do extremely stupid mistakes a lot more often than you would think, particularly in the heat of the moment!

[Pmalek]

I just saw an X post that claims that seedphrases generated on Coldcard Mk4s are beginning to be drained now as well. They were considered vulnerable already but the entropy was allegedly not as weak as the one on Mk3 with the affected firmware versions. Here is one such post.

Someone has also reported that they found a critical vulnerability in Bitkey. Perhaps there is a connection or AI found something.

[JiiBs]

it could be a cryptographic flaw within the hardware device where the device generates users seed phrases in such a way that an attacker could predict them, similar to the Ill Bloom vulnerability from earlier in the month.

This is the part that catches my interest, how possible can this be, it’s speaks highly of a possible vulnerability that I don’t wish to believe exists within the system. It sounds like an impossibility and could make users believe that the pathway to seed phrase generation isn’t safe.

The volume of Bitcoin that was stolen form this attack though, that’s a lot and in the millions of dollars, quite extreme.

[Mate2237]

If Coldcard knew about this (presumably), I'm thinking what could they have said to protect the victims? This is a critical vulnerability that targets cold storage users, something that hasn't ever occurred before, if I'm not mistaken.

I remember there was an Electrum vulnerability that directed the users to download a malware, but airgapped users were unaffected. This particular case must be unique, and I'm curious how the Coldcard team could have informed the victims without revealing to the attackers the source of the problem (so they could take advantage immediately.)

Good news for people who were drained - they appear to have found the identity of the attacker

Coldcard Wallet theft has an estimated rise to $70M

Galaxy Research has expanded its analysis of the Coldcard wallet incident, and identified 1,196 addresses that lost 1,082.65 BTC, which is approximately $70 million within a 41-minute time period.

Galaxy's on-chain analysis traced the losses across nearly 1,200 wallets, all affected during the same narrow timeframe, suggesting a coordinated exploit.

Source

[Italian Panic]

It seems the hacker left some traces along the way, such as paying for a blockchain service to check addresses with larger amounts, so it’s safe to say they had carefully studied where and how to strike. It also appears there was a limit of 200 transactions per wallet in the research method, leaving transactions beyond that limit untouched, this indicates a rudimentary approach to the software, which avoided wallets with overly complex transaction histories.

[gmaxwell]

In my locale, we always come to the conclusion in discussions that open source wallets are better than closed ones, obviously because they do not conceal surprises. But in the age of AI development in open source, if people don't see mistakes, AI will see them, which we witnessed (Coinkite wrote about the use of AI by intruders). If this vulnerability has not been found in the five years of its existence, it has given a false feeling that the code is perfect..
 Of course, the developers are also to blame for making the low-entropy decision. But I think that all developers with open source code should not disdain unnecessary code reviews, both by their own specialists and by the most intelligent AI...

This vulnerability would be just as fast to find with AI if closed source, maybe even faster.

If I were analyzing the firmware of a closed source wallet, I'd have the AI disassemble it and look for where the HW TRNG was read and trace from there.  In this case it wouldn't be read at all (or the code that did would be dead code), and so it would be obvious that it was vulnerable.

In a strange way the open source code was a distraction that hid the behavior because it mostly looked correct.  After the theft reports but before the cause of this was known someone I was talking to mentioned pointing AI at it and I said "Might want to disassemble the firmware to make sure a build issue hasn't made this code get used (link to the fallback PRNG)" and "Challenge it to show you in the disassembly".  --- it can be easy to get confused by misleading code but the binary is definitive.  AI makes analysis of the binary much easier.

A useful lesson to extract from this is that especially now that we have AI these devices should be both reviewed from a source code *and* binary basis, because some vulnerabilities (like this one) would be more obvious from the binary.  It's not an entirely new lesson-- e.g. in libsecp256k1 we've used binary level review (esp for things like timing sidechannels) but now its enough easier that it should be done regularly.

A good review approach would intentionally inject serious faults like this one, and make sure the review process catches it... with the hope that this implies if there were any unknown faults they'd be found too. With AI this could even be automated.

[NotATether]

I just saw an X post that claims that seedphrases generated on Coldcard Mk4s are beginning to be drained now as well. They were considered vulnerable already but the entropy was allegedly not as weak as the one on Mk3 with the affected firmware versions. Here is one such post.

Someone has also reported that they found a critical vulnerability in Bitkey. Perhaps there is a connection or AI found something.

OK, Coldcard as a brand is finished.

There's no coming back from this for them

And that Bitkey vulnerability better be fixed before it gets exploited too.

[Majestic-milf]

It seems the hacker left some traces along the way, such as paying for a blockchain service to check addresses with larger amounts, so it’s safe to say they had carefully studied where and how to strike. It also appears there was a limit of 200 transactions per wallet in the research method, leaving transactions beyond that limit untouched, this indicates a rudimentary approach to the software, which avoided wallets with overly complex transaction histories.

So let's say this will help in the investigation, how many users can recover their stolen Bitcoin. In relation to the ColdCard matter, River seems to benefit from this upset or vulnerability because over 3,679 Bitcoin have been recorded as an inflow. It appears that more people are now focused more on institutions managing their funds than trusting hardware wallets.

[free-bit.co.in]

This whole thread is scary to look. To think a bug can destroy your whole life worth of savings os really horrible. Everyone trusts hardware wallets a little more because they are the most secure wallets. But if a bug causes bitcoins to be drained from the wallet like this, then I don't know which hardware wallet to trust the most.

I feel very bad for those who have been affected by this. I hope that the theft of such a huge amount will definitely be caught.

You are right, we actually buy hardware wallet with expectation that it will never generate private key in vulnerable way. But this incident showed that just buying a hardware wallet doesn't automatically guarantee security. Implementation of seed generation is also very important. If you only update the firmware, it won't make an old seed secure if it was generated with a flawed implementation.

However, this particular case should not lead to conclusion that all hardware wallet are insecure.
At the same time, we should not trust any wallet blindly. No matter which wallet you are using it is always good to protect it with strong passphrase.