I’m talking about whether external control is required for software updates to occur. As with Windows, where Microsoft can replace software on your machine at any time. On Ubuntu, canonical can add backdoored software to their repositories but they can’t forcibly install software on your machine AFAIK. Maybe I just misunderstood the comment I was responding to.
e.g. law inforcement gets a tool "BreakEveryonesSecurity.exe/app" that can talk to any device they find. For that to work all devices have to be pre-compromised. Another work for pre-compromised is "insecure".
You trust your OS updates because your machine trusts the certificate it is presented with. Your machine trusts that certificate because it is signed by an authority which is in your machine's local list of trusted authorities.
If people mess up that list of trusted certificate authorities, it can easily compromise the system. Remember this?
https://en.wikipedia.org/wiki/Superfish#Lenovo_security_inci...
Now imagine if, instead of having an identifiable certificate authority which accidentally allowed third parties to intercept and modify your encrypted traffic, that were an undetectable feature of the encryption algorithm.
If that were the case, no matter how much you trusted the entity that understands the nature of the backdoor, you would not trust your OS.
An iOS device cannot install an update without an unlock, and any updates it does push are signed by an apple key, and does not make software with compromised security.
Apple then designs the system so that someone hacking Apple doesn't give them the ability to hack iOS devices. The way apple achieves this is to make the phones require a user unlock to update the device or to connect to another machine.
The only way to make magic software work would be for apple to deliberately compromise the device security. The moment they did that, that would mean there was a tool that would only need to be leaked once to compromise all apple devices. The solution is to not design a broken security model.
I assume the high end android devices are similar, but given android phones are still being shipped without a Secure Enclave I can imagine there are some trivially compromised devices.
So then let's get to how the update itself works.
The device gets a user unlock to occur (iirc if you have automatic updates enabled this means that your device has to have been unlocked in the last 24 hours?[1]). The device pulls an update from an apple server, verifies both the server it pulls from, and the bundle itself. No one else can publish or mitm the update. There's no magic software that can just be passed around - you would need to compromise apple itself, and compromise it to the point that you can get the updates signed by their production signing keys without anyone noticing.
If you're really unsure how all this works feel free to read their documentation: https://www.apple.com/business/site/docs/iOS_Security_Guide....
[1] I re-read the white paper's update mechanism and it sounds like it always requires an unlock before updating. So it's not a hysteresis. So automatic update is gated on an unlock before bed and asking permission to perform the update.
That's not quite applicable to this story though which seems to be just observing network traffic.