The recent Coldcard hardware wallet exploit, which has resulted in more than $100 million in stolen Bitcoin, is prompting many Bitcoin users to rethink one of the most fundamental aspects of self-custody: how private keys are generated.
In the aftermath of the incident, security researchers, developers, and long-time Bitcoin advocates are increasingly encouraging users to rely on physical sources of randomness, such as rolling dice, instead of trusting hardware wallets to generate wallet seeds automatically. The renewed interest reflects a broader shift toward independently verifiable security practices and reinforces Bitcoin’s long-standing principle: “Don’t trust, verify.”
How the Coldcard Vulnerability Emerged
The controversy centers on the way certain Coldcard hardware wallets generated the cryptographic entropy used to create wallet seed phrases.
Earlier versions of the device relied on the STM32 hardware True Random Number Generator (TRNG), which produces randomness using physical electronic processes. Hardware-based entropy is generally considered the foundation of secure wallet generation because it creates unpredictable private keys that are practically impossible to guess.
However, during a firmware redesign that began in March 2021, researchers discovered that newer Coldcard firmware versions no longer relied exclusively on the hardware random number generator.
Instead, they used MicroPython’s Yasmarang pseudo-random number generator (PRNG) under certain conditions.
Unlike a hardware TRNG, a pseudo-random generator produces sequences derived from mathematical algorithms. While suitable for many computing tasks, PRNGs are generally not considered sufficient on their own for generating highly secure cryptocurrency wallet seeds.
Security researchers concluded that this implementation dramatically reduced the amount of entropy available during wallet creation.
Reduced Entropy Increased Brute-Force Risk
According to technical analyses, affected Coldcard devices generated significantly less randomness than required for cryptographically secure Bitcoin wallets.
Coinkite estimated that:
- Coldcard Mk2 and Mk3 devices generated approximately 40 bits of entropy
- Mk4, Mk5, and Q devices produced around 70 bits of entropy
By comparison, a secure 12-word BIP-39 seed phrase typically requires 128 bits of entropy, while a 24-word seed uses 256 bits.
Although 40 to 70 bits of entropy remains a very large search space, it is dramatically smaller than industry security standards, making large-scale brute-force attacks substantially more practical for well-funded attackers using specialized hardware.
Researchers believe attackers have exploited this weakness to systematically recover wallet seed phrases and steal Bitcoin from vulnerable devices.
More Than $100 Million Stolen
Blockchain investigators estimate that multiple coordinated attack waves have already resulted in more than $100 million worth of Bitcoin being stolen, making the Coldcard incident one of the largest hardware wallet security failures in recent years.
Researchers continue monitoring blockchain activity, with some estimates suggesting losses could eventually exceed $130 million as additional compromised wallets are identified.
To better understand attacker behavior, Bitcoin developer James O’Beirne launched cktripwire, a monitoring project that tracks suspected attacker activity using specially created honeypot addresses.
The initiative aims to identify which wallet configurations remain vulnerable and determine how attackers prioritize compromised addresses.
Physical Entropy Returns to the Spotlight
The incident has revived one of Bitcoin’s oldest security recommendations: generate entropy yourself whenever possible.
Rather than relying entirely on a hardware wallet’s internal random number generator, many security experts now recommend supplementing—or even replacing—it with physical randomness.
The simplest method involves rolling ordinary six-sided dice dozens or even hundreds of times.
Each roll contributes additional randomness that users can independently verify without needing to trust proprietary hardware or firmware.
Because the randomness originates from a physical process that users directly observe, it removes one of the most critical trust assumptions involved in wallet generation.
Why Dice Are Becoming the Preferred Option
Dice-based entropy offers several important security advantages.
Unlike electronic random number generators, physical dice require no hidden firmware, embedded chips, or proprietary algorithms.
Every roll can be observed, recorded, and independently verified.
Even if a hardware wallet were compromised, externally generated entropy significantly reduces the possibility that an attacker could predict the resulting private key.
For many Bitcoin users, this transparency has become especially valuable following the Coldcard incident.
Multiple Methods Exist for Manual Seed Generation
Several techniques are now circulating throughout the Bitcoin community for generating wallet seeds using physical randomness.
Popular approaches include:
Dice Rolling
Rolling more than 100 six-sided dice generates enough randomness to create a secure 24-word BIP-39 seed phrase when combined with standard cryptographic hashing methods.
This remains one of the simplest and most widely recommended approaches.
Dice Lookup Tables
Some wallets, including BitBox, provide printable lookup tables that allow users to convert combinations of dice rolls directly into BIP-39 seed words without requiring electronic devices.
The method combines multiple dice rolls—and sometimes a coin flip to generate each word in the recovery phrase.
Codex32 and Debiasing Techniques
Advanced users have also adopted Codex32, which uses mathematical techniques such as Von Neumann extraction to eliminate bias from imperfect dice, ensuring the resulting entropy remains cryptographically secure even if the dice themselves are not perfectly random.
Paper Shuffle Methods
Another entirely offline approach involves printing every BIP-39 word onto identical slips of paper, thoroughly shuffling them, and randomly selecting the required number of words.
Commercial products such as Seedsticks and Entropia have simplified this process by providing standardized physical entropy kits.
Hardware Can Still Be Verified
Security experts emphasize that manually generating entropy does not eliminate the usefulness of hardware wallets.
Instead, it reduces their trust requirements.
After creating a seed phrase independently, users can verify that multiple hardware wallets derive the same:
- Extended public keys (xpubs)
- Wallet addresses
- Transaction signatures
Cross-checking multiple devices significantly increases confidence that no hidden manipulation has occurred during wallet generation.
Some advanced users also compare RFC 6979-compliant transaction signatures across different devices to detect potential attempts at secret data leakage.
Self-Custody Is Evolving, Not Disappearing
Although the Coldcard exploit has prompted renewed criticism of self-custody, many Bitcoin developers argue the incident demonstrates the importance of improving verification—not abandoning personal ownership.
The vulnerability did not expose a flaw in Bitcoin itself.
Rather, it highlighted the importance of independently validating every critical step involved in securing digital assets.
As wallet technology continues evolving, many users are now placing greater emphasis on transparent, auditable methods for generating private keys rather than relying solely on closed hardware implementations.
Security Begins With Trustworthy Entropy
The Coldcard incident has fundamentally reshaped conversations around Bitcoin wallet security.
While hardware wallets remain one of the safest methods for storing cryptocurrency, the exploit has reinforced that secure entropy generation is the foundation of every self-custody solution.
For an increasing number of Bitcoin holders, a simple pair of dice has become more than a randomization tool—it has become a symbol of verifiable security, reminding the community that true self-custody begins not with sophisticated hardware, but with independently verifiable randomness.
