Can you describe a scenario where Touch ID is safe against evil maid attacks (say, a chip is installed allowing anyone to transmit a certain signal that spoofs Touch ID) while also allowing unrestricted modifications by someone with physical possession of the device (as this T2 rooting post celebrates)?

Right now, that security is provided by Apple crypto-locking the Touch ID sensor to the T2 chip so that it cannot be modified to allow unauthorized access without being disabled altogether.

With the ability to bypass the restrictions of the T2 OS, that protection is stripped away, and replaced by .. nothing, as far as I can determine.

This is akin to removing your car’s electronic anti-theft system because it requires OEM keys. Sure, you can do so, but your car is a lot easier to steal, too. Only it’s not your car here, it’s your computer and all personal data on it, and all SSH keys you use to access remote servers, too.

I’m all for repairability but it’s worrying that the tech community is so invested in removing a padlock that offends them that they set aside security and risk issues in favor of rooting without addressing it at all. If this complete lack of interest in device security is the best we can do, we don’t deserve repairability, and we don’t deserve root.

I do not understand why you find it worrying that the tech community is invested in removing a restriction IN THE OPEN. If it can be removed for nefarious purposes, eventually, someone will do so and sell the exploit on the zero-day market (probably to a state actor). The existence of the vulnerability is just a fact, it is reality. Why do you feel safer not knowing about it? Whether you know about it or not the vulnerability is still there.

I do not ‘feel safer not knowing about it’. I feel safer knowing how to DFU a T2 Mac, though. It’s turning out to be very useful knowledge over time. I encourage you to learn how too.

Allow the users to install their own keys. Changing keys invalidates all encrypted/secured data. Which means you have to export the data if you do hardware changes and reimport it after supplying your own. Once you have your own keys installed you could sign additional hardware with them.

If apple is a viable root of trust then you yourself should be too. There's nothing magical that only apple can do.

I believe that Apple is burning their cryptographic key into readonly memory, so they would need to build out a readwrite pipeline and provide a secondary keystore option for "non-default" users that is writable by the hardware itself. That's a tall ask, but it's feasible, so we're good so far. The benefit to expert users with crypto competence is clear.

How would this benefit third-party repair shops, though?

The point is that you'd backup the keys in advance (when you initially set up the machine) and when the machine dies and your T2 is fried the repair shop can just replace it with a new T2, load the backed-up keys into it and give you back your machine with the data intact.

AFAIK, it's not just about the security of your data (you don't need a T2 chip to encrypt data), but also about discouraging theft of the hardware itself.

In light of that, how do you allow for components to be swapped out wholesale without breaking the security model?

Isn't the entire point that you can't just steal a Macbook, swap out the SSD, and now you have a functioning (stolen) laptop?

You can also get away with a much weaker passphrase if you can somehow rate-limit brute force attacks. As I understand it that's one of the things the T2 does, you can't authenticate except through the T2, and the T2 will slow you down and lock you out after n tries.

And I think that's actually incredibly underappreciated feature, most people are both bad at memorizing long, complex passwords/passphrases and not all that motivated since the threat is quite abstract. I've had a hard time convincing some people to have any sort of password at all on their laptop (that they travel with, work in all sorts of places and are overall pretty careless with), as it carries the risk of forgetting the password and it's cumbersome and inconvenient compared to being able to simply open the computer to a logged-in desktop.

Making security less of a hassle really does help with adoption. Having a short, relatively weak, but not trivial password is way way better than having none at all, and realistically that's all most people are ever going to have, so making those passwords hold up better is a really smart move that instantly ups the security of lots and lots of people out there.

But that's not something I believe you could do without a scheme like tieing the encrypted data to a complex secret inside a specific T2 that is uncracked, otherwise you could simply put the SSD in another computer and brute-force it there.

The initial transition from the apple root of trust to your own root (which you would then use to install new hardware) could require being authenticated, this way a thief couldn't do it while the legitimate owner could.

That's a route I hadn't thought of.

Seems like a great idea!

Trying to prevent theft that way is a fundamentally flawed approach. It's in the end all about controlling the phone you brought to prevent you from using it in any way they don't like or using it longer then they like (by repairing it).

Theft will happen anyway. You can even sell permanently-locked/bricked devices to people which doesn't look to closely at the sellers description. Sure you will need to sell them for cheap, but that's all.

That idea is like saying all cars must be always tracked, always link up with the drivers phone and be remote-controllable by there manufacturer to prevent theft.

Sure it would prevent theft, maybe, until people find ways to brake it. But it's still totally unreasonable with a lot of hidden cost to it.

E.g. in case of apple laptops the cost is losing a lot of small independent companies as well as any way to properly repair an Apple laptop. (Apple doe NOT provide proper repairs they at best replace whole components often the whole main board because it's one component when many damages tend to be similar because people use their devices similar and often are reasonable fixable with a bit of not-easy-but-not-very-hard-either soldering).

EDIT: And most important! The theft constraint can be archived to a reasonable degree WITHOUT locking out third party repair. A example (through not applicable to mac in this case) is how I setup my laptop with a custom EFI platform key/certificate and a BIOS password so to reuse it after theft people have to replace the BIOS chip soldered onto the motherboard (it has no publicly known master key or reset pin). Apple can archive similar things so that theft is more costy but third party repairs are still mostly unconstrained.

In general I agree with your points. But reducing the profit achievable on stolen devices can reduce the motivation to steal them. Whereas there are already big risks involved in stealing a car so adding DRM wouldn't be that much of an impediment in that case. (And cars already do use DRM in various ways.)