2 ms·

It would be entirely possible for the secure enclave to refuse a firmware update while the device is locked. Or as an alternative to refusing, it could ensure t

It would be entirely possible for the secure enclave to refuse a firmware update while the device is locked. Or as an alternative to refusing, it could ensure the NVRAM is reset in the process, rendering the encryption keys permanently lost.

ajross 11y ago
Arrgh, no What you suggest is not possible if the firmware for the enclave is stored off-chip in mutable storage. Or rather, you could do it, but you'd have to hash the firmware and simply refuse to run any new firmware without a data wipe. Things like eMMC's RPMB partitions are designed to work around this (by having the transactions encrypted by a secret key stored on the SoC, generally in wrote-once fuses), but then you have the problem that you'd have to wipe all data on any firmware update, which relevant to this thread we now know is not the case. Basically: people are hanging their determination that "the iPhone is secure from government snooping by technological measures" on a completely unattested fantasy design. The real device almost certainly doesn't work like that, and even Apple hasn't claimed that it does.
- sjwright 11y ago
  So what you're saying is it's possible, but your speculation is different to my speculation? I'm not sure it's useful to have a battle of hypotheses and assumptions couched in the language of confidence and certainty.
  - ajross 11y ago
    I'm saying that placing bets on the security of devices based on speculation is a bad idea. Lots of people seem to want to hide behind the "iPhone can't be snooped" shield that has been built in this community, and to my reasonably well-informed eyes that's misguided. But even on the technical side, my point was stronger: the fact that the enclave firmware was just updated to behave differently without wiping the data that it controls argues very (very!) strongly that the "snoop-proof" feature you believe exists... doesn't.
    - sjwright 11y ago
      I agree, we shouldn't be placing bets. I wasn't. I was just saying that it is technically possible to implement a secure enclave that can protect its data against firmware updates while the device is locked. The fact that the enclave firmware can be seamlessly updated while in an unlocked state is not evidence that the enclave firmware can be seamlessly in a locked state.

ajross 11y ago

Arrgh, no What you suggest is not possible if the firmware for the enclave is stored off-chip in mutable storage. Or rather, you could do it, but you'd have to hash the firmware and simply refuse to run any new firmware without a data wipe. Things like eMMC's RPMB partitions are designed to work around this (by having the transactions encrypted by a secret key stored on the SoC, generally in wrote-once fuses), but then you have the problem that you'd have to wipe all data on any firmware update, which relevant to this thread we now know is not the case. Basically: people are hanging their determination that "the iPhone is secure from government snooping by technological measures" on a completely unattested fantasy design. The real device almost certainly doesn't work like that, and even Apple hasn't claimed that it does.

sjwright 11y ago
So what you're saying is it's possible, but your speculation is different to my speculation? I'm not sure it's useful to have a battle of hypotheses and assumptions couched in the language of confidence and certainty.
- ajross 11y ago
  I'm saying that placing bets on the security of devices based on speculation is a bad idea. Lots of people seem to want to hide behind the "iPhone can't be snooped" shield that has been built in this community, and to my reasonably well-informed eyes that's misguided. But even on the technical side, my point was stronger: the fact that the enclave firmware was just updated to behave differently without wiping the data that it controls argues very (very!) strongly that the "snoop-proof" feature you believe exists... doesn't.
  - sjwright 11y ago
    I agree, we shouldn't be placing bets. I wasn't. I was just saying that it is technically possible to implement a secure enclave that can protect its data against firmware updates while the device is locked. The fact that the enclave firmware can be seamlessly updated while in an unlocked state is not evidence that the enclave firmware can be seamlessly in a locked state.