6 ms·
It's not reasonable to expect people to have Yubikeys. iPhone Keychain is about as good as it'll get realistically, and that somewhat relies on hardware securit
by hot_gril 3y ago
It's not reasonable to expect people to have Yubikeys. iPhone Keychain is about as good as it'll get realistically, and that somewhat relies on hardware security.
- xyst 3y ago"iPhone Keychain" - no thanks, I'll stick with a non-vendor specific provider. I am trying to escape that awful ecosystem, not dig myself further in.
- hot_gril 3y agoThe option of Yubikeys is fine as long as the basic 1P thing is painlessly usable too.
- overstay8930 3y agoyou know its trival to export, right? There's nothing more secure than Keychain if you're in the Apple ecosystem. Nothing gets more scrutiny from the entire industry, at least.
- recursive 3y agoThe "ecosystem" comes as non-several package. Like for instance, my pet issue "if" I'm in the ecosystem, I'd have to give up my headphone jack. And all the rest of it. The "if" is probably most of the problem.
- lxgr 3y agoIf you know of a way to export a passkey from iCloud Keychain to a non-Apple device, please do share it! Otherwise I'd call that lock-in as well.
- hot_gril 3y agoIt's easy on a Mac since Safari has a CSV export feature. No such thing on an iPhone.
- lxgr 3y agoHave you tried exporting a passkey that way? Last time I did, I only got passwords out, not passkeys (not that there is an interoperable standard for them anyway).
- hot_gril 3y agoOh, I didn't realize passkeys and totp aren't the same thing. Totp secrets go into the CSV. Don't think I even have any passkeys to test with. And supposedly 1Password doesn't let you export either. This seems bogus. I'd rather simply use a random per-site password; looks like passkeys are the same except non-interoperable.
- lxgr 3y agoBitwarden lets you export them as part of at least their JSON export, but unfortunately there's no specified interoperable format yet, so you can only import them back into Bitwarden (which you can at least self host; you could reimplement their serialization format if you're really determined). There's some movement in that area in the related FIDO working groups, but I think we'll (by design) never see something like CSV export, and it'll be more like a standardized account migration. > I'd rather simply use a random per-site password; looks like passkeys are the same except non-interoperable. They're significantly better than a random per-site password since they can't be compromised on the server side (due to being based on public key cryptography), unlike regular passwords and TOTPs.
- lukeschlather 3y agoActually I maybe misspoke and I might go further than that and say that services shouldn't be allowed to make any requirements about how hardware tokens work. This means if someone wants to use a software token that should be supported. And also I think this is why the passkey standard is bad, it sets rigid hardware requirements and the manufacturers will use this to drive planned obsolescence. If Apple and Microsoft have their way we will throw away $1000+ phones and laptops because someone found an exploit in the TPM that requires physical access.
- hot_gril 3y agoYes, that and WEI