Disassembled Trezor Safe 7 / Tampering Protection

Security
1

Hello there,

to be less susceptible to 5$ wrench attacks and be a bit safer while traveling, I would like my Trezor to look less like a Trezor, to draw less attention in case I’m robbed. I was thinking that I could tear apart the Trezor Safe 7, and only carry the mainboard and screen in my luggage, separately, without the frame. That would already look much less like a Trezor.

All parts besides the mainboard and touchscreen I would remove:

  • the frame
  • the battery
  • the UI FPC (power button and status LED)
  • the ANT FPC (NFC and wireless charging)

Some resulting questions:

  1. Would the Trezor Safe 7 be able to boot and operate with only the mainboard and touchscreen, when powered via USB? I dug a bit in the firmware code and had the impression that in theory that should be possible, even when the other components / their PCBs are not connected.
  2. I’ve seen that there seems to be a tamper detection (J10 in the schematic), which seems to be released when the board is not pressing against the screen anymore after removing it from the frame. Would it be possible to circumvent the tampering detection and use the Safe 7 without the frame? Would I have to bridge a connection between J10 and some other point, potentially J9? Or would it even suffice to simply press down J10?
  3. Would J7 and J6 also require some extra treatment?
  4. During disassembly, I will not be able to avoid triggering the tampering detection. Is it a recoverable state? From what I’ve seen, it seems to delete volatile state and show a warning on next boot, but not delete persisted secrets, seed, attestation keys or pairing keys of the various chips. After dismissing the warning and avoiding triggering J10 again, will the Safe 7 be able to boot again as a genuine Trezor device?

I’ve mostly researched the Safe 7. However, if the Safe 5 does happen to not have such tampering detection on the chip, that would be a viable option as well.

2

The tamper detection was in fact disabled because the spring loaded trigger didn’t work well enough and this was causing spurious RSODs.

In any case, current tamper detections halt the CPU but do not permanently brick the device.

From what my colleagues tell me, what you’re planning to do should be perfectly feasible, if, honestly, rather crazy.

You might also prefer to go for the Safe 5, simply because it’s much simpler internally. If you don’t need Bluetooth – which you shouldn’t, given that you’ll have to have the device wired anyway – it’s the more obvious choice.

3

How did you open it up?