Wallet Logo

Trezor Safe 3

Latest Release: 2.8.0 10th July 2024

Our wallet review process

We examine wallets starting at the code level and continue all the way up to the finished app that lives on your device. Provided below is an outline of each of these steps along with security tips for you and general test results.

Developer

Website Twitter Facebook Reddit

Custody

Private keys generated and held by user

As part of our Methodology, we ask: Is the provider ignorant of the keys?

The answer is "yes". Private keys are generated by the user on the wallet.
Read more

Source code

Public on github

Released

12th October 2023

Application build

The binary provided was reproducible from the code provided.
See the last Issue we created.

See test result
Tested 11th October 2024

Platform notes

There is no globally accepted definition of a hardware wallet. Some consider a paper with 12 words a hardware wallet - after all paper is a sort of hardware or at least not software and the 12 words are arguably a wallet(‘s backup). For the purpose of this project we adhere to higher standards in the hardware wallet section. We only consider a hardware wallet if dedicated hardware protects the private keys in a way that leaves the user in full and exclusive control of what transactions he signs or not. That means:

  • The device allows to create private keys offline

  • The device never shares private key material apart from an offline backup mechanism

  • The device displays receive addresses for confirmation
  • The device shares signed transactions after informed approval on the device without reliance on insecure external hardware

Passed all 12 tests

We answered the following questions in this order:

Is this product the original?

The answer is "yes".
If the answer was "no", we would mark it as "Fake" and the following would apply:

The answer is "no". We marked it as "Fake".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Fake" and the following would apply:

The bigger wallets often get imitated by scammers that abuse the reputation of the product by imitating its name, logo or both.

Imitating a competitor is a huge red flag and we urge you to not put any money into this product!

The product cannot be independently verified. If the provider puts your funds at risk on purpose or by accident, you will probably not know about the issue before people start losing money. If the provider is more criminally inclined he might have collected all the backups of all the wallets, ready to be emptied at the press of a button. The product might have a formidable track record but out of distress or change in management turns out to be evil from some point on, with nobody outside ever knowing before it is too late.
Can we expect the product to ever be released?

The answer is "yes".
If the answer was "no", we would mark it as "Announced but never delivered" and the following would apply:

The answer is "no". We marked it as "Announced but never delivered".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Announced but never delivered" and the following would apply:

Some products are promoted with great fund raising, marketing and ICOs, to disappear from one day to the other a week later or they are one-man side projects that get refined for months or even years to still never materialize in an actual product. Regardless, those are projects we consider “vaporware”.

Is this product available yet?

The answer is "yes".
If the answer was "no", we would mark it as "Un-Released" and the following would apply:

The answer is "no". We marked it as "Un-Released".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Un-Released" and the following would apply:

We focus on products that have the biggest impact if things go wrong and while pre-sales sometimes reach many thousands to buy into promises that never materialize, the damage is limited and there would be little definite to be said about an unreleased product anyway.

If you find a product in this category that was released meanwhile, please contact us to do a proper review!

Is it a wallet?

The answer is "yes".
If the answer was "no", we would mark it as "Not a wallet" and the following would apply:

The answer is "no". We marked it as "Not a wallet".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Not a wallet" and the following would apply:

If it’s called “wallet” but is actually only a portfolio tracker, we don’t look any deeper, assuming it is not meant to control funds. What has no funds, can’t lose your coins. It might still leak your financial history!

If you can buy Bitcoins with this app but only into another wallet, it’s not a wallet itself.

Is it for bitcoins?

The answer is "yes".
If the answer was "no", we would mark it as "A wallet but not for Bitcoin" and the following would apply:

The answer is "no". We marked it as "A wallet but not for Bitcoin".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "A wallet but not for Bitcoin" and the following would apply:

At this point we only look into wallets that at least also support BTC.

Is the provider ignorant of the keys?

The answer is "yes".
If the answer was "no", we would mark it as "Provided private keys" and the following would apply:

The answer is "no". We marked it as "Provided private keys".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Provided private keys" and the following would apply:

The best hardware wallet cannot guarantee that the provider deleted the keys if the private keys were put onto the device by them in the first place.

There is no way of knowing if the provider took a copy in the process. If they did, all funds controlled by those devices are potentially also under the control of the provider and could be moved out of the client’s control at any time at the provider’s discretion.

The product cannot be independently verified. If the provider puts your funds at risk on purpose or by accident, you will probably not know about the issue before people start losing money. If the provider is more criminally inclined he might have collected all the backups of all the wallets, ready to be emptied at the press of a button. The product might have a formidable track record but out of distress or change in management turns out to be evil from some point on, with nobody outside ever knowing before it is too late.
Does the device hide your keys from other devices?

The answer is "yes".
If the answer was "no", we would mark it as "Leaks Keys" and the following would apply:

The answer is "no". We marked it as "Leaks Keys".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Leaks Keys" and the following would apply:

Some people claim their paper wallet is a hardware wallet. Others use RFID chips with the private keys on them. A very crucial drawback of those systems is that in order to send a transaction, the private key has to be brought onto a different system that doesn’t necessarily share all the desired aspects of a hardware wallet.

Paper wallets need to be printed, exposing the keys to the PC and the printer even before sending funds to it.

Simple RFID based devices can’t sign transactions - they share the keys with whoever asked to use them for whatever they please.

There are even products that are perfectly capable of working in an air-gapped fashion but they still expose the keys to connected devices.

This verdict is reserved for key leakage under normal operation and does not apply to devices where a hack is known to be possible with special hardware.

The product cannot be independently verified. If the provider puts your funds at risk on purpose or by accident, you will probably not know about the issue before people start losing money. If the provider is more criminally inclined he might have collected all the backups of all the wallets, ready to be emptied at the press of a button. The product might have a formidable track record but out of distress or change in management turns out to be evil from some point on, with nobody outside ever knowing before it is too late.
Can the user verify and approve transactions on the device?

The answer is "yes".
If the answer was "no", we would mark it as "Bad Interface" and the following would apply:

The answer is "no". We marked it as "Bad Interface".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Bad Interface" and the following would apply:

These are devices that might generate secure private key material, outside the reach of the provider but that do not have the means to let the user verify transactions on the device itself. This verdict includes screen-less smart cards or USB-dongles.

The wallet lacks either a screen or buttons or both. In consequence, crucial elements of approving transactions is being delegated to other hardware such as a general purpose PC or phone which defeats the purpose of a hardware wallet. For big exit scams, a companion app could always request two signatures - one for the coffee you are paying and a second to empty your wallet completely. The former could be broadcast while the latter only gets collected for later use.

Another consquence of a missing screen is that the user is faced with the dilemma of either not making a backup or having to pass the backup through an insecure device for display or storage.

The software of the device might be perfect but this device cannot be recommended due to this fundamental flaw.

The product cannot be independently verified. If the provider puts your funds at risk on purpose or by accident, you will probably not know about the issue before people start losing money. If the provider is more criminally inclined he might have collected all the backups of all the wallets, ready to be emptied at the press of a button. The product might have a formidable track record but out of distress or change in management turns out to be evil from some point on, with nobody outside ever knowing before it is too late.
Is the source code publicly available?

The answer is "yes".
If the answer was "no", we would mark it as "No source for current release found" and the following would apply:

The answer is "no". We marked it as "No source for current release found".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "No source for current release found" and the following would apply:

A wallet that claims to not give the provider the means to steal the users’ funds might actually be lying. In the spirit of “Don’t trust - verify!” you don’t want to take the provider at his word, but trust that people hunting for fame and bug bounties could actually find flaws and back-doors in the wallet so the provider doesn’t dare to put these in.

Back-doors and flaws are frequently found in closed source products but some remain hidden for years. And even in open source security software there might be catastrophic flaws undiscovered for years.

An evil wallet provider would certainly prefer not to publish the code, as hiding it makes audits orders of magnitude harder.

For your security, you thus want the code to be available for review.

If the wallet provider doesn’t share up to date code, our analysis stops there as the wallet could steal your funds at any time, and there is no protection except the provider’s word.

“Up to date” strictly means that any instance of the product being updated without the source code being updated counts as closed source. This puts the burden on the provider to always first release the source code before releasing the product’s update. This paragraph is a clarification to our rules following a little poll.

We are not concerned about the license as long as it allows us to perform our analysis. For a security audit, it is not necessary that the provider allows others to use their code for a competing wallet. You should still prefer actual open source licenses as a competing wallet won’t use the code without giving it careful scrutiny.

The product cannot be independently verified. If the provider puts your funds at risk on purpose or by accident, you will probably not know about the issue before people start losing money. If the provider is more criminally inclined he might have collected all the backups of all the wallets, ready to be emptied at the press of a button. The product might have a formidable track record but out of distress or change in management turns out to be evil from some point on, with nobody outside ever knowing before it is too late.
Is the decompiled binary legible?

The answer is "yes".
If the answer was "no", we would mark it as "Obfuscated" and the following would apply:

The answer is "no". We marked it as "Obfuscated".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Obfuscated" and the following would apply:

When compiling source code to binary, usually a lot of meta information is retained. A variable storing a masterseed would usually still be called masterseed, so an auditor could inspect what happens to the masterseed. Does it get sent to some server? But obfuscation would rename it for example to _t12, making it harder to find what the product is doing with the masterseed.

In benign cases, code symbols are replaced by short strings to make the binary smaller but for the sake of transparency this should not be done for non-reproducible Bitcoin wallets. (Reproducible wallets could obfuscate the binary for size improvements as the reproducibility would assure the link between code and binary.)

Especially in the public source cases, obfuscation is a red flag. If the code is public, why obfuscate it?

As obfuscation is such a red flag when looking for transparency, we do also sometimes inspect the binaries of closed source apps.

As looking for code obfuscation is a more involved task, we do not inspect many apps but if we see other red flags, we might test this to then put the product into this red-flag category.

The product cannot be independently verified. If the provider puts your funds at risk on purpose or by accident, you will probably not know about the issue before people start losing money. If the provider is more criminally inclined he might have collected all the backups of all the wallets, ready to be emptied at the press of a button. The product might have a formidable track record but out of distress or change in management turns out to be evil from some point on, with nobody outside ever knowing before it is too late.
Can the product be built from the source provided?

The answer is "yes".
If the answer was "no", we would mark it as "Failed to build from source provided!" and the following would apply:

The answer is "no". We marked it as "Failed to build from source provided!".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Failed to build from source provided!" and the following would apply:

Published code doesn’t help much if the app fails to compile.

We try to compile the published source code using the published build instructions into a binary. If that fails, we might try to work around issues but if we consistently fail to build the app, we give it this verdict and open an issue in the issue tracker of the provider to hopefully verify their app later.

The product cannot be independently verified. If the provider puts your funds at risk on purpose or by accident, you will probably not know about the issue before people start losing money. If the provider is more criminally inclined he might have collected all the backups of all the wallets, ready to be emptied at the press of a button. The product might have a formidable track record but out of distress or change in management turns out to be evil from some point on, with nobody outside ever knowing before it is too late.
Does the published binary match the published source code?

The answer is "yes".
If the answer was "no", we would mark it as "Not reproducible from source provided" and the following would apply:

The answer is "no". We marked it as "Not reproducible from source provided".

We did not ask this question because we failed at a previous question.
If the answer was "no", we would mark it as "Not reproducible from source provided" and the following would apply:

Published code doesn’t help much if it is not what the published binary was built from. That is why we try to reproduce the binary. We

  1. obtain the binary from the provider
  2. compile the published source code using the published build instructions into a binary
  3. compare the two binaries
  4. we might spend some time working around issues that are easy to work around

If this fails, we might search if other revisions match or if we can deduct the source of the mismatch but generally consider it on the provider to provide the correct source code and build instructions to reproduce the build, so we usually open a ticket in their code repository.

In any case, the result is a discrepancy between the binary we can create and the binary we can find for download and any discrepancy might leak your backup to the server on purpose or by accident.

As we cannot verify that the source provided is the source the binary was compiled from, this category is only slightly better than closed source but for now we have hope projects come around and fix verifiability issues.

The product cannot be independently verified. If the provider puts your funds at risk on purpose or by accident, you will probably not know about the issue before people start losing money. If the provider is more criminally inclined he might have collected all the backups of all the wallets, ready to be emptied at the press of a button. The product might have a formidable track record but out of distress or change in management turns out to be evil from some point on, with nobody outside ever knowing before it is too late.

Application build test result

Updated Review 2024-10-11

Following the instructions enumerated below, we were able to successfully build version 2.8.0 of the Trezor Safe 3 firmware.

1. Download version 2.8.0 of the binaries from the official Trezor website:

$ wget https://data.trezor.io/firmware/t2b1/trezor-t2b1-2.8.0{,-bitcoinonly}.bin

2. Get the hashes of the binaries downloaded:

   $ sha256sum *.bin
   bd29cf141327fd0a0e04fff9213eeacb9adc273586dd8992ecb3b09d05320666  trezor-t2b1-2.8.0.bin
   7dfc875f9208524d00bcb8e56ef05e92f60e1a4bc94872dcabe01609a8177b35  trezor-t2b1-2.8.0-bitcoinonly.bin

3. Clone the repository for the trezor-firmware:

$ git clone https://github.com/trezor/trezor-firmware.git

4. Change directory to the cloned repository:

$ cd trezor-firmware

5. Checkout to version 2.8.0

$ git checkout core/v2.8.0

6. Run the trezor firmware build script

$ bash -c "./build-docker.sh --models R core/v2.8.0"

7. This resulted in the following fingerprints. Notice that Core-T, is not included:

   Built from commit a3a7b0cd7a8e0fe78726ddb1c3159c2972a7278a

   Fingerprints:
   5673e8cfc0f6cbae7a5e6c5b26473acdbda36a81d0fb074a680f4fba734690e1 build/core-R/bootloader/bootloader.bin
   cf3ce230a69a681199f74cf6ac8c6c431f8fa7e0d0183437f93c5cc029fbd155 build/core-R/firmware/firmware.bin
   5673e8cfc0f6cbae7a5e6c5b26473acdbda36a81d0fb074a680f4fba734690e1 build/core-R-bitcoinonly/bootloader/bootloader.bin
   ae088439d44fc8643b8de28e0d7a8720cd3dbb247619f2742604bbe884542558 build/core-R-bitcoinonly/firmware/firmware.bin

8. We proceed with signature zeroing

   danny@lw10:~/work/builds/hardware/trezorSafe3/trezor-firmware$ cp ../trezor-t2b1-2.8.0-bitcoinonly.bin trezor-t2b1-2.8.0-bitcoinonly.bin.zeroed
   danny@lw10:~/work/builds/hardware/trezorSafe3/trezor-firmware$ cp ../trezor-t2b1-2.8.0.bin trezor-t2b1-2.8.0.bin.zeroed
   danny@lw10:~/work/builds/hardware/trezorSafe3/trezor-firmware$ vendorHeaderSize=4608
   danny@lw10:~/work/builds/hardware/trezorSafe3/trezor-firmware$ seekSize=$(( 5567 - $vendorHeaderSize + 512 ))
   danny@lw10:~/work/builds/hardware/trezorSafe3/trezor-firmware$ dd if=/dev/zero of=trezor-t2b1-2.8.0.bin.zeroed bs=1 seek=$seekSize count=65 conv=notrunc
   65+0 records in
   65+0 records out
   65 bytes copied, 0.000164314 s, 396 kB/s
   danny@lw10:~/work/builds/hardware/trezorSafe3/trezor-firmware$ dd if=/dev/zero of=trezor-t2b1-2.8.0-bitcoinonly.bin.zeroed bs=1 seek=$seekSize count=65 conv=notrunc
   65+0 records in
   65+0 records out
   65 bytes copied, 0.000174494 s, 373 kB/s
   danny@lw10:~/work/builds/hardware/trezorSafe3/trezor-firmware$ sha256sum *.zeroed build/core-R{,-bitcoinonly}/firmware/firmware.bin | sort
   877aed88c703a89344ae9b098f84849a2e1db29c1740e71af5cc85042eeb8ec1  build/core-R/firmware/firmware.bin
   877aed88c703a89344ae9b098f84849a2e1db29c1740e71af5cc85042eeb8ec1  trezor-t2b1-2.8.0.bin.zeroed
   c1edd08f02b75430fbeedd77c4d155aadd81b030752fec4465698ab9a0b3d0e3  build/core-R-bitcoinonly/firmware/firmware.bin
   c1edd08f02b75430fbeedd77c4d155aadd81b030752fec4465698ab9a0b3d0e3  trezor-t2b1-2.8.0-bitcoinonly.bin.zeroed

After zeroing out 65 bytes after a specific offset, we find that the hashes of the standard official and built binaries match. The same can be said of the bitcoinonly binaries. We can assert that version 2.8.0 of the latest firmware available for the Trezor Safe 3 is reproducible.

Previous Review 2024-04-01

This is the latest model of the Trezor hardware wallets and it does feature a so called “secure element” or SE in short. With that, our first worry is if we have a firmware to review at all. Popular products like the Ledger Nano S    trade transparency for “security” by running their sofware on chips where the provider does not allow them to disclose the source, putting the provider in a position where any update could put funds at risk with no way of independent scrutiny.

As Trezor has been struggling with these SEs for a long time, to the point of working on their own SE that would allow scrutiny, while always working with truely open source, this product is probably doing the right thing, too. But … how can we verify?

It took quite long to find mentions of firmware verifiability for the Trezor Safe 3 but as metioned in a reply to this issue,

the Safe 3 firmware is the same type and format as the TT so the exact same instructions apply.

the only change (which we do need to document) is that the path to the TS3 binary is build/core-r/fimware/firmware.bin, and similarly for the model T the path changes to core-t

This is great but if for example the keys get created by only the SE, we are back at having to trust its unknown code, right? So how exactly do they use the SE?

This article has answers.

We do not run code on the chip itself. The Secure Element simply stores a secret that can be used to decrypt the recovery seed, i.e., it never actually knows what your recovery seed is.

This is good. It implies that the SE neither generates the secret, following the same path as for example the BitBox02    .

So, where is the firmware to be found? Quite a long search doesn’t yield a binary to download and their chatbot appears to not understand us neither.

In the repo where we can find the firmware for their other products, we can see cryptic descriptors t1b1, t2b1 and t2t1 which according to this code translate as follows:

TREZOR_ONE = T1B1
TREZOR_T = T2T1
TREZOR_R = T2B1
TREZOR_SAFE3 = T2B1
TREZOR_T3T1 = T2B1
TREZOR_DISC1 = DISC1
TREZOR_DISC2 = DISC2

So … T2B1 is the right one. Let’s see …

$ wget https://data.trezor.io/firmware/t2b1/trezor-t2b1-2.6.4{,-bitcoinonly}.bin
$ sha256sum *.bin
3940dc0615c651104baf0e10147550d4ad2e44e2ef317a94ed36245e3e016bf2  trezor-t2b1-2.6.4.bin
203d5d8d50ced8d75086c418ba2ee4cb9b2857df27821767c935922db4f30184  trezor-t2b1-2.6.4-bitcoinonly.bin
...
$ git clone https://github.com/trezor/trezor-firmware.git
$ cd trezor-firmware
$ git checkout core/v2.6.4 
$ bash -c "./build-docker.sh --models R core/v2.6.4"
...
Fingerprints:
7558899d0e9a551738c5e29b4b27741d00a472fc8f9e47eddbfc983fc236cddd build/core-R/bootloader/bootloader.bin
5ac16cb5002aa607908be376378a7fd1a1bc18f7b05e7a047cb1365840cc93ef build/core-R/firmware/firmware.bin
7558899d0e9a551738c5e29b4b27741d00a472fc8f9e47eddbfc983fc236cddd build/core-R-bitcoinonly/bootloader/bootloader.bin
013d595fc621c12324afd90721c6a37d055d853f6af54d5432e27e6a425656dd build/core-R-bitcoinonly/firmware/firmware.bin
3302cba4ab90b667aec6049d2299ff08fa13613beb624b338f3275fee04aaf7a build/core-T/bootloader/bootloader.bin
441faa92156e8ae0b8247f9434c3ec8cf6ffd872f16fc593b22c4460dfd93913 build/core-T/firmware/firmware.bin
3302cba4ab90b667aec6049d2299ff08fa13613beb624b338f3275fee04aaf7a build/core-T-bitcoinonly/bootloader/bootloader.bin
e78da8a00354dd1223da081600f881b71bd297dd565e7a2c0a9880e52575d127 build/core-T-bitcoinonly/firmware/firmware.bin
$ cp ../trezor-t2b1-2.6.4-bitcoinonly.bin trezor-t2b1-2.6.4-bitcoinonly.bin.zeroed
$ cp ../trezor-t2b1-2.6.4.bin trezor-t2b1-2.6.4.bin.zeroed
$ vendorHeaderSize=4608
$ seekSize=$(( 5567 - $vendorHeaderSize + 512 ))
$ dd if=/dev/zero of=trezor-t2b1-2.6.4.bin.zeroed bs=1 seek=$seekSize count=65 conv=notrunc
$ dd if=/dev/zero of=trezor-t2b1-2.6.4-bitcoinonly.bin.zeroed bs=1 seek=$seekSize count=65 conv=notrunc
$ sha256sum *.zeroed build/core-R{,-bitcoinonly}/firmware/firmware.bin | sort
ac2aadad8850f6d23763e543ddd2604e5760eeb7c7ec747f1c1015246125207c  build/core-R-bitcoinonly/firmware/firmware.bin
ac2aadad8850f6d23763e543ddd2604e5760eeb7c7ec747f1c1015246125207c  trezor-t2b1-2.6.4-bitcoinonly.bin.zeroed
c723a55315ed5528db602d8ef0eebeb4a8a9ed96a0e237122398a999fab5e75e  build/core-R/firmware/firmware.bin
c723a55315ed5528db602d8ef0eebeb4a8a9ed96a0e237122398a999fab5e75e  trezor-t2b1-2.6.4.bin.zeroed

This binary is reproducible except for the signature which is to be expected.

Tests performed by Daniel Andrei R. Garcia, Leo Wandersleb

Previous application build tests

1st April 2024 2.6.4  

Disclaimer

Our Analysis is not a full code review! We plan to make code reviews available in the future but even then it will never be a stamp of approval but rather a list of incidents and questionable coding practice. Nasa sends probes to space that crash due to software bugs despite a huge budget and stringent scrutiny.

Do your own research

In addition to reading our analysis, it is important to do your own checks. Before transferring any bitcoin to your wallet, look up reviews for the wallet you want to use. They should be easy to find. If they aren't, that itself is a reason to be extra careful.