Hacker News Reader

Comment by JoeAltmaier | Hacker News Reader

That's the whole issue, isn't it? You can't create software that operates only on a single instance of a computer. It'll run on all of them.

View on HN

That's not the case – if Apple produce a signed iOS version that allows the passcode retry limit to be bypassed on a single device, then in what meaningful sense does it operate on another device?

> if Apple produce a signed iOS version that allows the passcode retry limit to be bypassed on a single device, then in what meaningful sense does it operate on another device?

What prevents the iOS image from being loaded onto another device?

Apple isn't magic; the code they write to verify device identity isn't going to be the first perfect, unbuggy, unexploitable code written in human history. if(device_udid == terrorist_id) {...} might seem infallible, to you, but the reality is that the device_udid is just SHA1(Wifi MAC + Bluetooth MAC + ECID + Serial). All of those are writable, some via the Baseband and some via physical access. Generating SHA1 collisions is completely feasible for ~$1,000,000 of computing time, which is chump change to nation states.

There is no infallible way for Apple to make an iOS version for one single device.

Edit: And stories like http://abcnews.go.com/Technology/york-da-access-175-iphones-... make it absolutely clear that this is not stoping at a single device.

I guess when the FBI compels them again. Or once they have a foot in the door, compels them to remove the signing requirement, or wants their own ability to sign.

This is not mysterious. Can the FBI require a software company to subvert their own security guarantees? Should they? Do bank vaults have back doors for the FBI?

I guess when the FBI compels them again

That's a bit of a specious argument though – of course, if they produce different software in the future, it could be used again!

If the FBI want their own ability to sign Apple updates, then that's obviously a different situation. But the point you were making – that it's not possible to write software that will run on only one device – is not true in this case.

This is not mysterious. Can the FBI require a software company to subvert their own security guarantees? Should they? Do bank vaults have back doors for the FBI?

I don't really know the right answer.

Apple could write code that checks the hardware ID supplied and sign only that version. It's been sufficient to prevent hacking so far and is the major protection relied upon until the latest generation, so I don't think this is an issue at all.

The issue is the precedent that is made. Once they create the special firmware, even if it is locked to one device, they have stated that law enforcement have already told them they have hundreds more devices they want unlocked. They would have a much easier time going to court and compelling Apple to resign the existing special firmware they made to work on these other hundreds of devices. Apple could no longer argue they had to create something to fulfill the warrant, since it was already created.

From what I can tell the UDID lock is not actually secure enough and is not designed to be a critical path around the encryption algorithm.

An earlier comment with more technical detail: https://news.ycombinator.com/item?id=11141499

UDID lock is good enough for tying development builds to specific devices but is not an unbreakable guarantee the software cannot be run on another device.

This comment simply makes claims like altering data in transit inside the architecture of the CPU.

Yes, if you can do that, then you can flash anything to the phone. When someone shows me that happening, then I'll believe in this technology.

No, I proposed 3 different ways of hacking the UDID in that comment. 1) disrupting the read of ECID in transit, 2) re-writing ECID by hacking the BPP, 3) a SHA1 hash collision. Any one of the 3 will do.

Even then you cannot "flash anything" to the phone. But you can flash a build signed by Apple which hard-codes a UDID check.

Wouldn't these mechanisms apply to key verification? How exactly can you argue that you can modify data in realtime inside the phone's architecture, but couldn't possibly modify data in realtime inside the phone's architecture?

Only certain parts of the hardware responsible for key storage and verification are architected to be resilient against physical and electronic attack, but the entire phone is not.

For example changing the WiFi or Bluetooth MAC is not locked down and effects the UDID. Because the UDID was never intended to be a way to bypass the device encryption it was not designed with anywhere near the same level of care and sophistication as handling the encryption key itself.

> For example changing the WiFi or Bluetooth MAC is not locked down and effects the UDID. Because the UDID was never intended to be a way to bypass the device encryption it was not designed with anywhere near the same level of care and sophistication as handling the encryption key itself.

I never suggested the UDID. Please don't ascribe a claim to me that was never made. SHA512 the serial and macs independently. Job done.

...and they give it to the FBI who then changes that ID as they please? Remove signing altogether? Its the fear of a blank check that drives this issue.

No – iOS is signed, and changing the ID would invalidate the signature.

Taking the private key from Apple with a warrant, national security letter, subpoena, or other methods is much easier than creating a custom version of iOS.

It is foolish to assume that the FBI would stop at this special version of iOS, especially given how they have been arguing and fighting to break encryption for over two decades.

edit, on icebraining's interpretation:

That's exactly right - once the difficult part about targeting an individual phone is finished, it's easy to take the iOS signing key and target any device you want.

> Taking the private key from Apple with a warrant, national security letter, subpoena, or other methods is much easier than creating a custom version of iOS.

Clearly it is not, otherwise they would have done this. They did not.

I think it's meant taking the key used to signed that custom version of iOS. That is, if the FBI can force Apple to write the custom version, they can also coerce the key from them to sign their own copies, preventing your suggested blocking mechanism.

This is what I understood, at least, from the parent's post.

And can Apple guarantee that the ID can't be changed inside the device? If they can't it makes a lot of sense for Apple to refuse creating such a tool.

The ID can be easily changed. A quick Google search will easily confirm this.

The question is can you change one devices ID to exactly match another. This might require hacking the BPP or finding a SHA1 hash collision, both of which are absolutely possible but not necessarily trivial.

Pretty much, as far as I'm aware; the device ID cannot be changed. Bear in mind that this entire discussion is applicable only to older iPhones, in any case.

That guarantee until the latest generation was the only thing protecting the hardware. It has never been altered or hacked that I am aware of. Doing so would require modifying the physical construction of the chips on the device.

Possible? Maybe, but if it was that simple the FBI could simply change the signing key. It's not, and they can't.

Reply on news.ycombinator.com