Skip to Content
DocumentationSecurityHardware and research

Hardware and research

This is the deeper part of the security section. Start with what a hardware wallet does, signing and scams, and offline setups if those topics are new to you.

Here the questions are what the hardware protects, which parts can be inspected, and what the evidence says about particular attacks. No single component answers all three.

Why KeepKey does not use a secure element

Does KeepKey have a Secure Element?

Production KeepKey devices use an STM32F205 general-purpose microcontroller without a separate secure element. The firmware coordinates key operations, the display, the physical button and USB communication.

A secure element is a specialized chip designed to resist attacks on sensitive data and operations. It can add meaningful resistance to probing, fault injection and extraction. Its effectiveness depends on the chip, its configuration and the surrounding system. Apple’s Secure Enclave is a different implementation of isolated security processing; the names are not interchangeable.

KeepKey’s design favors accessible components, reviewable firmware and a practical path for people to inspect, build and maintain the system. The tradeoff includes less dedicated hardware resistance to physical extraction. Physical custody and, where appropriate, a strong passphrase are important parts of that choice.

Open hardware and reproducible firmware

Public firmware lets researchers inspect how keys and requests are handled. Reproducible builds let people compare compiled artifacts with an authenticated reference. Public schematics and documented components help builders examine and reproduce a design.

The firmware source  and DIY project  provide starting points. The DIY project uses its own off-the-shelf hardware and firmware target; an educational build should not be assumed to have the physical protections of a production device.

Specialized chips can introduce purchasing restrictions, provisioning, tooling or supplier-specific interfaces. KeepKey’s current direction favors transparent parts and practical multi-vendor sourcing. More domestic sourcing is a design goal, rather than a description of every current component.

Commodity microcontrollers still have suppliers and implementation details that must be trusted. A public datasheet is not a transistor-level verification of silicon. Buying directly from a manufacturer can reduce some risks, but does not establish that a device has never been intercepted or modified upstream.

Go deeper: compare open and closed trust boundaries

Secure elements differ. Publishing wallet firmware does not automatically open every chip, and adding a secure element does not automatically close every part of a wallet.

  • Trezor Safe 7  combines TROPIC01, OPTIGA Trust M and an STM32U5 microcontroller, with cooperating protection layers.
  • Ledger’s explanation  identifies substantial published software and restricted low-level operating-system components tied to its chip supplier agreement.
  • TROPIC01 distribution  shows that accessible secure elements are becoming available. Evaluate documentation, provisioning and supplier dependence for the particular part.

These describe different review boundaries and dependencies, rather than an overall security ranking.

Physical attacks and BIP39 passphrases

Kraken’s 2019 research  demonstrated extracting encrypted recovery material from a KeepKey through voltage glitching, followed by an offline attack on PIN protection. The report recommends physical custody and a BIP39 passphrase. Its historical equipment and timing estimates are not benchmarks for every current attacker or device.

A BIP39 passphrase combines with recovery words to derive a different wallet. KeepKey does not persist it in device storage. Reading the stored recovery material alone is therefore insufficient to derive that wallet: an attacker must also obtain or guess the passphrase.

Episode MULTI 05 · Passphrase wallets · 1:25 · KeepKey, step by step

Use an unpredictable secret and protect a separate backup. Every passphrase produces a wallet, so a typo can open a different, empty wallet without an error. Verify that you can reproduce the intended wallet before funding it.

BIP39 fixes the derivation at 2,048 PBKDF2-HMAC-SHA512 iterations . Once recovery words are known, guesses can be tested offline. Strength depends on unpredictability, not a character count alone. Changing the iteration count changes the derived wallet.

The passphrase and derived secrets must be processed during use. A compromised host, altered firmware or implant can capture them then. A passphrase is not universal protection against compromise or coercion. Use the passphrase guide for setup and recovery responsibilities.

Firmware verification

Verification asks several different questions: who published a file, whether its bytes match the expected artifact, and whether a build from source matches that reference. A checksum from the same untrusted download location does not establish publisher identity.

Episode BTC 08 · Verify your firmware · 1:30 · KeepKey, step by step

The firmware verification workflow walks through authentication, preparing a complete build environment, comparison and installation checks. Matching bytes does not prove the source has no bugs. Device-reported information also depends on the device reporting honestly.

For ordinary updates, follow the firmware update guide. Use release status and the Test Atlas to distinguish published artifacts, candidate tests, emulator captures and physical-device evidence.

Has KeepKey ever been “hacked”?

Researchers have reported software and physical-access vulnerabilities in KeepKey. Assess the affected versions, prerequisites and resolution of each finding. A firmware update addresses particular software issues; it does not remove every hardware attack class or revoke an attacker’s existing copy of secrets.

Independent research, audits and AI-assisted review

A commissioned independent audit has a named reviewer, defined scope, tested versions and a report. A public disclosure documents a particular finding. Internal tests and AI-assisted review can help find bugs, but do not constitute an independent audit.

Under current ownership, KeepKey has not commissioned a new traditional third-party security audit. Historical audits do not establish the security of current releases. Independent researchers’ disclosures, internal testing and automated review—including hosted and self-hosted AI tools—provide different kinds of evidence. Findings still require reproduction, technical review and appropriate regression checks.

Vulnerability history

These are selected original reports, rather than an exhaustive inventory. Historical fixed versions identify where a particular issue was addressed; follow the current update guide for installation instructions.

FindingWhat the research demonstratedDocumented resolution or limit
CVE-2019-14355: OLED side channel Display power consumption can expose information about screen contents.The research covers several products. Check mitigation for the specific device; CVE-2019-14353 identifies Trezor One, not KeepKey.
CVE-2021-31616: stack buffer overflow USB or WebUSB requests to an unlocked vulnerable device could expose secrets, including a cached passphrase.Introduced in 7.0.3 and patched in 7.1.0, according to the researcher.
CVE-2022-30330: supervisor flaws Malicious firmware could bypass privilege boundaries and alter protected flash, including the bootloader.The researcher documents bootloader bl_v2.1.4, released with firmware 7.3.2, as the fix.
CVE-2023-27892: memory exfiltration Crafted Ethereum messages could expose memory on the display of a physically accessible, unlocked device.The report identifies 7.5.2–7.6.0 as affected and 7.7.0 as fixed.
2019 physical-extraction research Voltage glitching extracted encrypted recovery material for an offline attack on PIN protection.Physical custody and a strong passphrase address different parts of this hardware risk. A software patch does not establish that the whole attack class is removed.

Read the firmware security advisories , release notes  and Kraken technical follow-up  with these examples. A public CVE count is not a loss-rate study or an overall security score.

If a device or backup may be compromised

Treat loss, suspected modification or exposed recovery material as a security incident. Use a trusted device and environment to assess moving funds to a newly generated wallet. Restoring the same recovery words on another device does not revoke an attacker’s copy.

Use support for help without sharing recovery words or passphrases. Report suspected firmware vulnerabilities through the firmware security reporting instructions .

Return to: Security basics, or choose a practical Bitcoin workflow.

Last updated on