Yes, hence the potting. You'd have to figure out a way to open the device without breaking it.
If need be, just grind and cut till you are back to the one silicon wafer that contains crypto keys, then rebuild the whole lot from scratch, add your evil parts, then re-pot it.
What you’re describing is a different thing. It’s an “evil maid” attack, when you tamper with a wallet that is already set up by the user. Also, it would be ultra difficult and expensive.
What you’re saying is possible. But:
It would probably be cheaper to build a total fake Ledger Wallet, then just replace the new real wallet and deliver a fake with a weak entropy generator. Any case, that’s not what has happened this time.
(The MITM supply chain attack which has happened succeeds because it preserves the original secure hardware but intercepts the secret data at the display interface. But a low-effort mimic of the entire device like I suggested? That wouldn’t work).
And as this workaround leaves the original Secure Element intact, good luck detecting the supply chain attack.
Man can make it, man can break it.
The best I've done is to modify a potted chip by very careful thermal imaging to figure out where the part was that blocked the feature I wanted to unlock (market segmentation) and then drill through the case to a depth just higher than the chip. You'd have definitely been able to spot the difference. This was at the millimeter level. Recently there was a post about a group doing this at the chip level, which I thought was extremely impressive, and again, even with the recipe out there duplicating it would still be very hard.
https://www.xda-developers.com/how-drill-defeated-the-xbox-3...
Sounds like a good approach when you only have a single (or just a few) of the devices so you need to be careful.
With hardware wallets (aka substantial incentive), then the attackers could order a bunch of them for sacrificial zero-ing in the exact places to drill with a CNC/microscope/etc.
Yes, initially I had just the one. But once I had it figured out there were lots of other people that wanted their device 'improved' so that jig got used plenty of times. Even so working on other people's hardware always felt very scary, especially poking around with a drill in a sensitive piece of hardware. Even so, they all worked, it was a neat little trick.
But it was clear what it did: as soon as the channel selector went to a forbidden channel the blob sprung into action to disable the chip underneath. I measured the effect on the pins underneath that were accessible from the other side of the board and noticed that a whole bunch of them got shorted to ground. That was the hint I needed. A friend ran a laser game company and they had a nice little IR camera with built in display, that was used for calibration of the IR beam collimators in the laser guns (the laser doesn't actually do any of the shooting, the thing is basically just a very powerful remote control with a lens that sends out a relatively narrow beam). I set it up above the the blob and set the channel selector to a good channel and let it warm up. Had a look with the camera. Then set the channel selector to a bad channel, let it warm up again and had another look with the camera. The difference was pretty obvious, there was a hotspot right in the middle of the blob. I marked the spot and made a drill jig, then it took me a few days to gather courage and stick the drill in, it worked on the first try. That gave 18 nearly empty channels to play with. Once I had it figured out the second one was a lot easier :)