I have to support some non technical people in our company once in a while. Trust me, this stuff is way too hard for normal people. People I know that own things like a ubi key are almost without exception IT professionals that know what they are doing. Outside IT professionals in IT departments, I don't know a lot of people that have much awareness of this stuff.
The trend in this space is to not have dependencies on dedicated hardware and use multi factor via phone apps and other things. It works, it's easy to explain to users. And lots of apps do this now. Google authenticator is still used but is increasingly positioned as a backup option to more user friendly alternatives.
WebAuthn is not a great success so far. It launched with a lot of fanfare a few years ago and then nothing much at all happened. In it's current form, that's not going to change any time soon.
Today there is no migration path from one MacOS WebKit browser (Safari) to another (which also has an implementation of PassKey), since the third party browser can't share the same authenticator key.
As you say, all of this is far too complicated for an end user - my suggestions to avoid the problem and migrate only work for tech savvy users that know every site they use. That's maybe 0.1% of users.
My worry if we see adoption of passkeys is that only 0.1% of the browser market becomes contestable, as everyone else is stuck locked into their incumbent browser with no way to enroll their new browser's passkey into the sites they use.
That means anyone can implement it, including a physical token (which can be entirely open source, like the solokeys dongle).
The real concern here is friction for the "average end user" - passkeys IMHO are a net-good thing, as long as we don't see this result in everyone regressing towards "single factor auth" in some way. As it stands though, WebAuthn/Passkey gives you a level of phishing resistance that ought to raise the bar on compromising accounts.
The part I do fear about third party password managers is that they'll potentially end up lowering the level of security that WebAuthn heralded, by normalising pure "software" authenticators - putting TOTP seeds into bitwarden alongside passwords feels like putting all your eggs in one basket, even if it's a reasonably good basket. A physical WebAuthn key gives you a level of hardware isolation (limited attack surface, time-bounded attack surface, physical contact required per-authentication) that will be lost if everyone moves to software-based tokens.
On the other hand, if people are replacing a globally re-used password with a "passkey", it's a lot better. If they are replacing a hardware token with a software token, that's a small step backwards. If most people are still manually using the same password everywhere, it's probably a net step forwards.
It is not possible to take anything Apple or Google do in this area to be in good faith.
If that happens it will be 100% the fault of those specific passkey implementations. There's no reason why exporting a passkey database should be any more difficult than exporting a password database.
That isn't to say I disagree with you though.
This could all be mitigated with a little bit of tooling (allowing an existing passkey to enrol a new passkey from another device), which would also help users of hardware tokens to potentially create a way to auto-enrol an off-site key.
I do believe WebAuthn is a good-faith attempt to get away from the pervasive problem of "use the same useless password everywhere", but it makes a range of compromises which (intentionally or otherwise) create a level of cryptographic vendor lock-in that I don't think many people have recognised yet.
Or any other mismatched devices ecosystem, the locked in ecosystem dream of apple is only applicable to extremely tiny amounts of people.
You shouldn't get locked in with a USB hardware token - you can enrol it in Chrome on one computer, and then authenticate via Firefox in another computer.
The issue is really the software-based "passkey" implementation. As long as you're not solely reliant on one company for login (i.e. you enrol multiple keys, one of which is portable and interoperable, AKA a hardware token) you can safely add software-based ones for convenience without getting locked in - you can always use the hardware token to get back in and enrol a new device.
Before "cloud keychain" (i.e. software-backed, like Apple's implementation), your only "safe" option was to have 2x hardware tokens, and try to keep one off-site, but still accessible enough you could enrol it on all the sites you use.
I fear with "cloud keychain" Webauthn, we are heading for a world where getting locked out of (or banned from) your "FAMNG" account will lock you out of everything else, to an extent we've not really seen before - no access to your synced keychain secret will prevent you from logging in to everything. Physical tokens remaining interoperable should give technical users an insurance policy against this, but without some kind of portability (i.e. pairing hardware webauthn keys), I fear it won't be practical enough to keep users safe, and independent of the keychain providers.
You inherently can't copy-paste a passkey, since it's an asymmetric public/private keypair authentication. These keys are (usually) decrypted by a single symmetric key that you protect well. If you allow that key to be exported, you're back to "one password for every website"!
When this is done on a hardware-protected security engine (which doesn't permit any extraction of the key), it's arguably quite secure, but you don't then have any migration path.
Does Firefox use the keychain - on macOS it does not even try
However, the lock-in concern remains: how can you export passkeys to another password manager the same way you can export passwords to it?
Chrome and iOS will pop up a QR code for webauthn over BLE or whatever it's called (there's also caBLE[0]) whenever you're either enrolling or trying to log in via webauthn. This means that, as long as you have access to any device with your old passkeys, you can log in on your browser/android phone.
0: https://blog.millerti.me/2021/06/18/previewing-chromes-cable...
1 ) Hardware authenticators won't spit their root secret, almost by design.
2 ) Webauthn doesn't require that the authenticator store the list of accounts, also by design (for privacy). So if you want to switch from iPhone to Android, you have to remember all websites you used Passkey on, and go one by one hunting down the right security settings page.
The specific challenge here is around software webauthn for passwordless access (think using Safari to create an account on a site). In this scenario, the average user has no portable authenticator. They cannot move to a new browser - you install Chrome, but can't log in from Chrome, as only safari can do a passkey login.
Even if chrome supports an equivalent setup (their version of passkey over Google sync, for example), you can't enroll it - to enroll, you need to sign in using Safari. To enroll your new device (chrome), you need to use it. You can't get logged in on chrome to do this. The average user has no option. A tech savvy user could manually copy session cookies to steal their own session, perhaps, or use a hardware key as a "bridge".
In essence, if you sign up for something using a passkey, you won't be able to easily leave that ecosystem at all, without pretty advanced tech knowledge (using a dedicated hardware webauthn key, or stealing and porting session cookies).
My separate observation about a lack of support for hardware keys to be "paired" to support an off-site backup use-case is unrelated, but perhaps relevant for tech savvy users who want to better "own" their own identity, and link their webauthn keys together for backup use-cases. Otherwise you have to maintain a list or spreadsheet of every site you use - I have one, so I can ensure I enroll each token I have with each service!
I think people who evangelize Webauthn need to carefully convey the risks and remind everyone that end users need backups (multiple authenticators, backup codes...). Hopefully, down the road, it will force interoperability between big manufacturers so one authenticator can authorize another for all websites in one go (this probably requires websites to have a standard way to enroll new authenticators).
> My separate observation about a lack of support for hardware keys to be "paired" to support an off-site backup
This is worrying me more. Interoperability between tech giants is bad but the sovereign solution may never get there.
Say you install an app/visit a website on your phone and register an account with a passkey. My current understanding is that on iOS 16, the passkey lives in your iCloud keychain. If you want to sign in on a Mac on Safari, you can just visit the website and the discoverable credential from your phone will appear when you try to log into the site with webauthn. The website will be able to tell that you're logging in from a new device and optionally require additional authentication.
If you want to sign in on a device that doesn't have access to your keychain, you can use your phone as an authenticator over a combination of Bluetooth and a tunnel server by scanning a QR code on that device with your phone. The site is then supposed to prompt you to register the new device with whatever its local passkey solution is.
The best source I could find for how this protocol works at a technical level is an episode of "Security. Cryptography. Whatever" on passkeys. I guess the specs aren't exactly public yet (at least since I last checked) and are only available to fido alliance members.
I've been trying to work on figuring out ways to build a "passkey manager" of sorts to live up to the potential webauthn offers with hardware-backed credentials that are also synced and backed up (to an offsite key). As far as I can tell, as mentioned in another comment, this just doesn't seem to be a priority for the fido alliance, which is a real shame.
I'm cautiously waiting to see how 1Password deals with passkeys, given that they're one of the few FIDO members with a vested interest in being cross platform, but I'm betting they'll just implement a software keystore built into their current vaults without any hardware backing.
If you look at the APIs available on MacOS/iOS at present [0], and how WebAuthn works, you need to "trust" the client-side software to correctly identify the site origin (i.e. protocol, domain, port) and pass it to a physical security key. In the case of a software token, you need to restrict which app(s) can pass this information through to the system carrying out authentication.
The end result right now, as I can see it, is that if you sign up to a website using a "passkey" via Safari (for example), there is no real supported workflow that isn't site-specific, for migrating to a new browser - you are effectively beholden to Apple's keychain for getting you back into that site.
If you want to enrol a new device (say an Android phone using its own software implementation of WebAuthn), you would need to find a way to sign into your old account (which requires your MacOS/iOS passkey to log in), then enrol your Android device (which is a separate physical device). You'll likely get stuck here.
An expert user could add a physical FIDO2 token to their account from the old device (Safari), and use that as the authenticator to log in on the new device, but this is significant friction for a non-technical user.
While moving users away from passwords might make sense, I do think we'll see these "passwordless" authentication mechanisms become the next major point of friction in user browser choice. Even if you can implement WebAuthn in your rival browser, unless you get access to the system key store, you won't be able to let users sign in. That will prevent them switching.
[0] https://developer.apple.com/documentation/xcode/supporting-a...