The recent incident with Coldcard hardware wallets has revealed a critical gap in approaches to independent testing of devices of this class. Kraken's Chief Security Officer Nick Percoco pointed to a fundamental problem: standard checks cannot guarantee that the random number generator claimed by the manufacturer is actually used in the working firmware.
The essence of the problem is that auditors can only confirm the presence of an approved random number generator (TRNG) in the hardware. However, they are unable to verify that the final firmware, which creates the seed phrase, actually utilizes this component. As Percoco emphasized, users are essentially being asked to trust the implementation of a critical security function by the manufacturer without any independent verification of the full entropy generation path.
The root of the problem and the scale of the damage
Let me recall the timeline. The manufacturer Coinkite admitted that an error embedded in the code since March 2021 led to the use of a weaker MicroPython pseudo-random number generator (PRNG) instead of the hardware TRNG. A technical analysis showed that due to build configuration peculiarities, the fallback Yasmarang generator was used. For a number of models, this meant a lack of cryptographic entropy, while for others it critically reduced the search space to four bytes.
The consequences turned out to be catastrophic. Based on my estimates, derived from analyzing fund movements, the total damage has already exceeded $90 million. This involves several waves of attacks: from a massive withdrawal of 1082 BTC in 41 minutes to subsequent transactions of hundreds of bitcoins. Notably, for older models (Mk2, Mk3), the entropy of seed phrases was only about 40 bits, making their compromise a matter of time even without sophisticated equipment. For newer models (Mk4, Mk5, Q), the figure was higher — about 72 bits — but still far from the safe 128 bits.
The need to revise standards
Coinkite has halted shipments and destroyed remaining devices with vulnerable firmware, but this does not solve the systemic problem. Percoco rightly draws parallels with other industries: payment terminals are not allowed into operation without independent laboratory testing, and cryptographic modules for US government agencies undergo strict verification of entropy sources according to NIST SP 800-90B and BSI AIS-31 standards.
The hardware wallet industry desperately lacks similar regulation. Existing certifications — Common Criteria, CSPN, or secure element audits — often check individual components but not the end-to-end path from the randomness source to the executable code. This incident is not just a story about one manufacturer's negligence but a wake-up call for the entire ecosystem. Until a mechanism for mandatory verification of actual entropy usage in working firmware is implemented, such risks will remain a "time bomb" for crypto asset holders.
My recommendation: users, especially those with large sums, should not only update firmware but also demand transparency from manufacturers regarding entropy sources, as well as use additional protective measures, such as hardware wallets with open-source code and independent full-stack verification.