@glyph I’m curious what you think of Apple’s recovery processes. They have recovery contact and legacy contact (provide access upon death). They both use two key shares (one with Apple, one with the chosen contact).
https://support.apple.com/guide/security/account-recovery-contact-security-secafa525057/web
https://support.apple.com/guide/security/legacy-contact-security-secebf027fb8/web
@glyph These were actually added last year in iOS 15, although without much fanfare. I think they were in preparation of E2EE iCloud. You have to enabled a recovery contact or recovery key (28 alphanumeric code) for Advanced Data Protection.
A flow chart would be really nice though. And I don’t think they have warnings for if you don’t enable ADP, although I haven’t made a new Apple ID in a long time so not sure.
As for your earlier SMS 2FA vs YubiKey lockout, I’m less familiar with but that could still be an issue. My understanding is if you have a logged in device, you’ll get notified of login attempts by a prompt on your trusted devices asking for approval which will supply the 6-digit 2FA code. If you don’t have a device, you can request a code via SMS instead. Not sure how this works if you enabled FIDO keys instead; the documentation just says you can be locked out if you lose all keys and trusted devices.
Someone on Reddit tested various scenarios if you’re interested: https://www.reddit.com/r/apple/comments/10mgr5l/security_keys_2fa_account_recovery_testing/
@glyph I think there is an issue around differentiating threat models.
My great grandma has a low odds of a concentrated attack and fewer online accounts with less info, by two orders of magnitude than most people. SMS 1/2fa is probably most secure thing she can do, most effective, and appropriate for her circumstances.
So much of infosec, (1) focuses only on worst case scenarios and the highest threat models, (2) desires perfection, and (3) have access to systems that will be targeted.
@glyph yubikeys make sense for them, not for most people (usability matters!)
Re perfection: this is an engineering thing (guilty as charged), you want the things you create or are vested in to be flawless.
@glyph Sure. I've seen that too.
I just think the threat modeling is part of the gap. You can put in a middle ground between grandma and infosec.