The blast radius is narrower than the technique suggests however. There is no remote component, so the attacker must be physically at the keyboard and know the right combination of keys. The machine must also have at least one account linked to an online Microsoft account, because the reset flow that anchors the attack depends on it.
Devices running only local accounts are not affected.



It does something very different than you’d naively expect, and is doing that just fine.
TPM just makes sure the code you run wasn’t manipulated/changed.
Or commonly on other device categories to make sure the code the manufacturer put there wasn’t changed by you to give you access.
In either case, if there are bugs you can exploit without changing code, then you get the same access as the operating system.
So if you are on a shitty OS like microslops windoes you should set a password query as early as possible, since that minimizes the chances. Ideally using a bios password or a wrapper decrypting the disk before the OS loads at all. Since microslop keeps adding both stupid bugs in early boot, and also actual (presumed) backdoors in e.g. the recovery.
Meanwhile on e.g. Linux, using TPM can essentially give you disk encryption protection but with the normal login screen and even multiple users.
That’s secure boot, not TPM.
When you hook it into disk encryption, it’s tpm.
Very much strongly related though.
Just like secureboot, things are checksummed, and the sum is then handed to tpm. Tpm will advance its state according to the things it is passed in this way. When it comes to decryption, the tpm will give out a key that depends on all those previous inputs.
If any step of the chain was different, the key will be different.
Of course, if you manage to exploit some of the boot process and take control without the component entering a new checksum, you can then add the remaining ones tpm would have normally gotten and get the unlock key.
Similar to secureboot, where you could then make it run any other code and execute any other .efi file without that being verified.
Difference is that with just secureboot, if I put the disk in another system with secureboot disabled, or turn of secureboot in the bios, I can just run my own code.
But with tpm, if I put the disk in a different system the tpm there won’t have the same key.
And if I say disable secureboot then the bios would still checksum the first .efi binary, and if it matches then that would be programmed to checksum and submit the next, etc. so without any bugs in uefi I should be unable to insert my own boot file while not ruining the tpm state with the checksum of that file.
And of course disabling tpm would equally leave me without ability to get the key.
I thought TPMs worked more like RPMb? Isn’t the key physically stored in some memory region, but it just refuses to give it out unless the inputs are cryptographically correct?
I didn’t think it was re-deriving the Bitlocker key each time?