The recent incident with Coldcard hardware wallets, which led to massive fund losses, has highlighted a fundamental problem in the approach to security verification of such devices. Kraken exchange's Chief Security Officer Nick Percoco pointed out a critical gap: standard audits verify the presence of a certified random number generator (TRNG) in the chip, but do not guarantee that the device's actual firmware uses this specific component.
The essence of the problem is both simple and frightening: users are asked to trust the manufacturer in implementing the most critical function—seed phrase generation—without independent confirmation that the declared entropy path is actually executed in the working code. It's like checking that a safe has a top-tier lock installed, but not checking that the door actually closes.
The Root of the Vulnerability: 40 Bits Instead of 128
Let me recall the timeline. The manufacturer Coinkite warned about the issue on July 30, about 30 hours after the first suspicious mass fund withdrawals. The error traces back to March 2021, when a new cryptographic library was integrated into the seed creation process. Instead of the hardware TRNG, the code began using a weaker pseudo-random number generator (PRNG) from MicroPython.
A technical analysis conducted by Block engineers showed that the production configuration defined the MICROPY_HW_ENABLE_RNG parameter as zero, and the libngu library only checked for the presence of the macro, not its activity. As a result, for Mk2 and Mk3, cryptographic entropy in ngu.random was completely absent. For the newer models Mk4, Q, and Mk5, the situation was better but still unsatisfactory: the entropy added by the secure element was limited to four bytes.
The outcome was catastrophic. For Mk2 and Mk3, the actual entropy of seed phrases was about 40 bits instead of the expected 128. For Mk4, Mk5, and Q—approximately 72 bits. This is a search space that can be brute-forced using specialized hardware.
Scale of Losses and Industry Reaction
Losses have already exceeded $90 million. Galaxy Research analysts tracked several waves of attacks: the first wave—1196 addresses and 1082.65 BTC, then two more, including a fresh attack on 448.7 BTC. Notably, Coinkite halted shipments and destroyed the remaining affected devices, but the firmware update does not fix already created seed phrases. Users are advised to migrate to new wallets, but to do so without haste to avoid creating additional risks.
The incident raises the question of the need for stricter standards for hardware wallets. Unlike the payment industry or cryptographic modules for government agencies, where laboratory testing of entropy sources is mandatory (under NIST SP 800-90B or BSI AIS-31 standards), the cold wallet industry relies on the manufacturer's integrity and audits they pay for themselves. Clearly, this is not enough.
My comment: This incident is not just a story about a bug in a specific model. It's a wake-up call for the entire industry. We trust hardware wallets with the storage of billions of dollars in assets, but the very system of their certification remains immature. Until mandatory independent audits that verify the full entropy path—from the generator to the executable code—appear, users will remain vulnerable to errors that are not visible at first glance. The question is not whether the next vulnerability will be found, but when it will happen.