Hacker News Reader

Comment by londons_explore | Hacker News Reader

Pretty disappointing that simply shorting two pins at the right time is sufficient to bypass security on a modern machine by a reputable vendor.

This eeprom, being security relevant, should be signed by the TPM, read into RAM, and if the signature check fails then the machine should consider itself stolen.

The only recovery method should be to replace the EEPROM contents with another valid image, also signed by the TPM. And that image should be available from the manufacturer for anyone able to show they are the device owner.

View on HN

You can kind of do that, although the process is a bit inverted.

BIOS settings are "measured" into the TPM, into a set of "platform configuration registers" (PCRs). The TPM itself takes no position on what the "correct" value of the PCRs is; however, it is possible to bind data against the PCRs. For instance, you can bind your disk encryption keys against the PCRs. Then, if the eeprom changes, the PCRs will be different, and the TPM will refuse to give you the disk encryption key.

The hardware itself is still valuable in the sense that a common thief cares about, but all of your data is protected, and you are not any more risk of using a compromised system then if the attacket simply swapped the unit for one that looks similar.

> The hardware itself is still valuable in the sense that a common thief cares about

Anti theft protection needs to make the hardware unusable when triggered sadly... Apples iCloud lock effectively does this - locked iDevices sell on the black market for only about 20% of what an unlocked device sells for - and that 20% represents the values of the fraction of components that are not serial number locked.

Signing with the TPM doesn't work super well, given that the TPM's inherent keys can't be used for signing as a privacy protection mechanism. You'd need to store an encrypted key in the image that could be used for signing purposes, but what stops an attacker from simply wiring up the TPM directly, generating a new key pair, and storing that instead? This is a legitimately difficult problem to solve well.

> but what stops an attacker from simply wiring up the TPM directly, generating a new key pair, and storing that instead

Simply?

Did you just equate resetting a bios password with wiring up a TPM and generating a new key pair?

The situation we're discussing already involves taking the machine apart to short the security eeprom during boot. Once you've accepted physical attacks as part of your threat model it seems reasonable to say that any mitigations you're designing should be resilient to physical attacks as well.

TPM is nonsense, i can trivially replace the processor/TPM module w/ whatever I wish.

If it was used to encrypt a disk, then what does replacing TPM accomplish exactly?

if the disk is encrypted, then the bios password does nothing at all. The access to the BIOS is a separate issue entirely.

Why are we even discussing TPM if it is not being used for its main purpose?

>modern machine

10 year old, last lenove gen where this was possible

Reply on news.ycombinator.com