Trezor Hardware Wallets and How Self-Custody Actually Works

What a Trezor is, and what it is not

Physically, a Trezor is a palm-sized unit with a display, either two buttons or a touchscreen, and a USB port. It has no battery and no network connection of its own. Plugged in, it draws power and speaks to a companion application; unplugged, it is an inert object that reveals nothing about the wallet it protects.

It is worth being precise about storage, because the popular phrasing is misleading. A Trezor does not contain your coins. Balances live on their respective blockchains and always have. What the device holds is the secret that proves you are allowed to move them: a master seed from which every address and every private key in your wallet is mathematically derived. Lose the device and you lose nothing, provided the backup is intact. Lose the backup and the device together, and the coins are unreachable forever.

The second thing a Trezor is not is a guarantee for everything you do online. It protects a key. It cannot stop you approving a malicious smart contract, sending funds to an address you were tricked into copying, or reading a convincing email and typing your recovery words into a form. The device narrows the attack surface dramatically. It does not remove your judgment from the loop.

Not your keys, not your coins. The old exchange-era adage is the entire product brief, compressed into six words.

Where Trezor came from

SatoshiLabs announced the first Trezor in 2013 and began shipping the following year. The founders, among them Marek Palatinus and Pavol Rusnak, had been running Bitcoin mining and pool infrastructure and wanted a way to hold coins that did not depend on the hygiene of a desktop operating system. The result, later renamed Trezor One when a second model arrived, established the pattern that almost every hardware wallet since has copied: keys generated on the device, transaction details shown on a screen the host computer cannot repaint, and confirmation by a physical press.

The same people were deeply involved in the standards the whole industry now runs on. BIP-39, the specification behind the familiar list of recovery words, was co-authored by Trezor's creators, and the company publishes its own series of SatoshiLabs Improvement Proposals. Trezor Model T, released in 2018, added a color touchscreen so the PIN and the seed could be entered on the device rather than through the computer, and introduced multi-share backup based on the openly published SLIP-39 scheme.

Later devices moved secrets into a dedicated secure element while keeping the firmware open. Trezor Safe 3 arrived in 2023, followed by Trezor Safe 5 in 2024 with a color touchscreen and haptic feedback. Across more than a decade the company has kept firmware, hardware design files and standards work in public, which is a large part of why Trezor is treated as a reference implementation by researchers, by rival manufacturers, and by wallet developers who need something to test against. The encyclopedia entry on the device is a reasonable neutral starting point for the company history.

How a Trezor protects a private key

When you set up a Trezor, the device generates a master secret using its own random number generator, mixed with entropy supplied by the host as a hedge against a weak on-device source. From that secret it produces a recovery seed of 12, 20 or 24 words, depending on the model and the backup type you choose. The words appear on the device screen and nowhere else. They are never sent to the computer, which is why setup refuses to advance until you have confirmed a few of them back.

The seed is a compact encoding of a very large number, and the standards known as BIP-32 and BIP-39 turn that number into an unlimited tree of key pairs. A derivation path identifies one specific address inside that tree. Your Trezor can hand out extended public keys so the desktop application can watch balances and construct transactions, while the private half of every pair stays inside the hardware.

ILLUSTRATIVE DERIVATION (BIP-84 NATIVE SEGWIT)

  seed words        24 words shown on device display only
  master key        m
  account path      m/84'/0'/0'
  first address     m/84'/0'/0'/0/0
  exported to host  extended PUBLIC key (xpub/zpub)
  never exported    any private key, the seed, the passphrase

Signing is where the design earns its keep. The companion application assembles an unsigned transaction and passes it to the Trezor over USB. The device parses it, displays the amount, the fee and the destination address on its own screen, and then waits. Nothing is signed until you press the button or tap confirm. Malware on the computer may propose whatever it likes, but it cannot forge that display and it cannot obtain a signature you did not authorize.

Access to a Trezor is gated by a PIN entered on a scrambled keypad, so a keylogger watching your mouse positions learns nothing useful. Wrong guesses trigger an exponentially increasing delay, which turns brute force into a project measured in years rather than minutes. The PIN protects the device sitting in your drawer. It is not what protects the words you wrote on paper.

One further design decision matters more than it sounds. A Trezor exposes a deliberately narrow interface: it accepts a defined set of message types, answers with signatures and public data, and does nothing else. There is no general file system to attack and no command that politely asks for the seed. Getting the secret out therefore requires either your cooperation or physical laboratory work, which is the subject of the next section but one.

The Trezor lineup compared

Four devices are commonly encountered. Trezor One is the long-running entry model with two buttons and a monochrome screen. Trezor Model T added a color touchscreen and a microSD slot. Trezor Safe 3 brought a certified secure element to the compact two-button shape, and Trezor Safe 5 pairs that secure element with a color touchscreen and haptic feedback.

The meaningful split is between the older pair and the Safe pair. Trezor One and Model T keep secrets in the flash memory of a general-purpose microcontroller, protected by the PIN and, if you set one, a passphrase. The Safe devices add an EAL6+ certified secure element that guards the PIN and seed material behind hardware built to resist physical extraction, and that also allows the device to prove to the application that it is genuine rather than a counterfeit.

Coin coverage differs as well. Every model handles Bitcoin, Ethereum and the large universe of ERC-20 tokens. Trezor One does not support several chains that need newer cryptography or more memory, Cardano, Monero, XRP and Tezos among them, so confirm the specific asset you intend to hold before buying on price alone.

Trezor Suite and the software side

Trezor Suite is the desktop and browser application that pairs with the device. It shows balances, builds and broadcasts transactions, organizes accounts per coin, and delivers firmware updates. The application is open source, and the desktop build can route its traffic through Tor so that your addresses are not tied to your home connection by whichever server it queries for blockchain data.

A few features earn their place and are routinely overlooked: labeling for accounts and individual transactions, coin control for Bitcoin so you choose which unspent outputs to spend, and a watch-only view for checking balances without the device attached. Trezor Suite also surfaces buy, sell and swap partners. Those are third-party services with their own terms and identity requirements, and using them is optional in every sense.

You are not locked into the first-party application either. Because Trezor implements published standards, third-party wallets such as Electrum, Sparrow and MetaMask can drive the device directly, which matters if you want a particular coin-control workflow, a multisignature setup, or an Ethereum interface you already know. A companion mobile app covers portfolio checking when you are away from your desk.

How Trezor holds up against real attacks

Threat modeling is the honest way to talk about a Trezor. Against remote attackers, meaning malware, a hostile browser extension or a breached exchange, the device is extremely effective, because the key never touches the compromised machine and every outgoing payment has to survive a look at an independent screen. Against an attacker who holds your device and has time in a laboratory, the answer depends on which model you own.

Researchers have demonstrated physical key extraction from the models without a secure element. Work published by Kraken Security Labs in 2020 used voltage glitching to pull the encrypted seed from Trezor One and Model T units, and in 2023 the firm Unciphered publicized a comparable recovery from a Model T. Both required possession of the hardware, specialist equipment and real expertise, and both are defeated by a strong passphrase, which is never stored on the device at all. The secure element in Trezor Safe 3 and Trezor Safe 5 exists precisely to raise the cost of this class of attack.

Supply chain is the other physical concern. Every Trezor ships in tamper-evident packaging, firmware is signed, the device warns loudly if unofficial firmware is present, and the Safe models can cryptographically attest that they are authentic on first connection. Buy from the manufacturer or an authorized reseller. A cheap Trezor from an anonymous marketplace listing is the single most common way people end up with a wallet whose recovery seed somebody else already wrote down.

In practice, though, most losses involving hardware wallets are social rather than technical. Fake Trezor support pages, cloned applications, browser pop-ups claiming your firmware is compromised, and urgent messages asking you to validate your wallet all funnel toward the same request: type your recovery words. In early 2024 the company disclosed that a third-party support ticketing portal had been accessed without authorization, exposing contact details of users who had written in. No funds or keys were at risk, but the episode is a clean illustration of why contact lists are valuable to phishing crews in the first place.

Openness is part of the security argument rather than a marketing flourish. The firmware and much of the hardware design for Trezor are published, builds can be reproduced independently, and outside researchers have repeatedly found problems and had them fixed in public view. A closed device may well be just as secure, but you would have to take that on faith rather than verify it.

Recovery seed, passphrase and multi-share backup

The recovery seed is the wallet. Anyone holding those words controls the funds, in any compatible software, permanently, without needing your Trezor at all. That is why the words show up only on the device display, why you write them on the supplied card or stamp them into metal, and why photographing them or saving them in a password manager quietly converts a hardware wallet back into a hot one.

A passphrase is an optional extra string combined with the seed to derive an entirely separate wallet. Each different passphrase produces a different set of accounts, and nothing on the Trezor records which ones exist, so there is no list for an attacker to find. It defends against the physical extraction attacks described above and against someone stumbling onto your written seed. It also has no recovery path whatsoever: forget the exact string, including spacing and capitalization, and that wallet is gone.

Multi-share backup, the SLIP-39 scheme introduced with Trezor Model T and supported on the Safe models, splits the secret into several shares with a threshold, three of five for instance. A single share reveals nothing on its own, and losing one is survivable. It suits people spreading backups across locations or trusted relatives, at the cost of a procedure that is harder to explain to whoever eventually has to follow it.

Whatever scheme you pick, test it. Wipe the device and restore from the backup while the amount at stake is still small, so you find out early that your handwriting is legible and your process works. Steel plates survive fire and flood in a way that paper cannot, and geographic separation is what protects you from the single unlucky event that destroys both your Trezor and the card sitting in the same drawer.

Trezor compared with other ways to hold crypto

Custody choices are a spectrum, not a ranking. An exchange account is convenient and outsources both the security work and the risk to a company. A software wallet gives you the keys but keeps them on a machine that also runs a browser and downloads files. A Trezor keeps the keys offline and inserts a confirmation screen into every transaction. Multisignature arrangements, often built from several hardware wallets, remove the single point of failure altogether in exchange for real complexity.

The comparison that matters most is against a phone or desktop wallet, since that is what most people are upgrading from. Both derive keys from a seed and both can sign transactions. The difference is that a Trezor cannot be read by malware running on the host, and it shows you the transaction on a display that host cannot control. For any amount you would be genuinely upset to lose, that separation is worth the price of the hardware and one careful evening of setup.

A short Trezor timeline

The dates below mark the releases that changed what the device could do, rather than every firmware revision along the way.

1. 2013

   The first Trezor is announced

   SatoshiLabs unveils the concept of a dedicated offline signing device for Bitcoin.

2. 2014

   Trezor One ships

   Two buttons, a monochrome screen, and the template every later hardware wallet follows.

3. 2018

   Trezor Model T

   Color touchscreen, on-device seed and PIN entry, and SLIP-39 multi-share backup.

4. 2023

   Trezor Safe 3

   A certified secure element and device authenticity checks in the compact button form factor.

5. 2024

   Trezor Safe 5

   Secure element plus a color touchscreen with haptic feedback.

How to set up a Trezor

A new Trezor takes about twenty minutes to commission if you do not rush the backup. Do it at a desk, alone, with a pen already in hand, and with no phone camera pointed at the table.

1. Step 01

   Buy from an official source and inspect the box

   Order from the manufacturer or a listed reseller, then check the packaging and seals before anything else. If the box looks opened or the device arrives already showing a seed, stop and contact support.

2. Step 02

   Install the app and the firmware

   Connect the Trezor, install Trezor Suite from the official download, and let it verify and write the current firmware. Devices ship without firmware deliberately, so this step is normal rather than a warning sign.

3. Step 03

   Create the wallet and record the backup

   Let the Trezor generate a fresh wallet, then write the recovery words down in order on the supplied card or on metal. Do not type them into anything. Complete the on-device check that confirms you copied them correctly.

4. Step 04

   Set a PIN and decide about a passphrase

   Choose a PIN you will actually remember, since there is no reset beyond restoring from your backup. If you are protecting a meaningful amount, read how passphrases work before enabling one and make the choice deliberately.

5. Step 05

   Test with a small amount, then test the recovery

   Send a small sum, verifying the receiving address on the Trezor screen rather than in the browser. Then wipe the device and restore from your written backup. Only once the restore has succeeded should you move the rest of your holdings.

Living with a Trezor over the long term

Day to day, your Trezor spends most of its life unplugged in a drawer, which is exactly right. Firmware updates arrive a few times a year and are worth installing, because they carry both new coin support and security fixes. The update process never asks for your recovery words, so any prompt that does is an attack, no matter how convincing the page around it looks.

Plan for the case where you are not around to explain any of it. Someone will need to know that a Trezor exists, where the backup lives, and whether a passphrase is required to see the real balance. Written instructions stored separately from the seed itself are the usual answer, and having a trusted person walk through them once, on a test wallet, is the step almost everybody skips.

Trezor questions people actually ask