Skip to main content


Hacker friends:

Am I correct in thinking passkeys replace "something you know" with something you have," or am I missing something here?

[lots of good discussion in replies, but in summary: don't, not even once.]

This entry was edited (1 month ago)
in reply to Michael Lucas

@Like Mucas :flan_yikes: As I understand it - they replace "something you know" with "something in your password manager" or "something you store on a USB device" which they are hoping is the same as "something you have", and I really don't think it counts.
in reply to silverwizard

its not something you store on a USB device, the USB device does a cryptographic computation every time it's used so the device is actually something you have. It's not a file.
in reply to silverwizard

well yes there are keys but they never leave the device, and they are supposed to be inaccessible even when you have access the physical key (there was a vulnerability discovered in older yubikeys that made them copiable recently, but if someone has access to the usb key physically they could just steal it at this point as with any "something you own")

But sure, there are keys in it. But following this logic everything in our universe is information so everything you own could technically be copied, but that's not a good way to think about it is it

in reply to Michael Lucas

yes, but you also need to authenticate on that device. passkey on phone - auth on phone. passkey in windows/mac , you need to be authenticated / enter pin / fingerprint / security key
in reply to Eugen

@ieugen so more like a passphrase-encrypted SSH auth key, it sounds like.
in reply to Michael Lucas

@ieugen …that is generated per website und unique to that Site and you don’t have to know how it works and you can’t export it (you can sync to $cloud provider though)
in reply to Michael Lucas

@ieugen yes, but with a unique key for each account. You can even use a ssh key as a passkey: github.com/bulwarkid/ssh-passk…
in reply to Michael Lucas

@ieugen As far as I can tell, this is all they are. Not the same as a yubikey's OTP generation.
in reply to Michael Lucas

something with an under-developed underlying spec, inconsistently implemented, usually tied to something a vendor knows. this is not the “hardware key” implementation you’re looking for. in my opinion. rip my mentions.
in reply to sungo

@sungo

Your opinion matches my understanding of what I've read.

I shall join you in rip-dom.

in reply to Michael Lucas

@sungo right there with you two. The more I read the spec docs for passkeys, the clearer a picture I have of the concerns and priorities of the companies pushing them, and the more convinced I become that passkeys are For Someone But Not For Me™.

I believe @mhoye coined the phrase "what if you wanted your SSH privkey to have a landlord"

in reply to sungo

@sungo 3FA: something you can lose, something you can forget, and something we can implement poorly.

reshared this

in reply to Michael Lucas

ime they replace "something you know" with "something a vendor has"
in reply to Michael Lucas

Indeed. think of it like a… asymmetric key-based authentication on SSH.

The spec however is not polished yet. They should’ve just made it simple and done, instead of complicated and incomplete.

in reply to Michael Lucas

Unhelpful joke
My understanding is they replace it with "don't worry your pretty little head, daddy Google will look after that for you."
in reply to Michael Lucas

ish, but there's also two things that somewhat got dumped into a single marketing term.
Assuming physical device, like a yubikey, there are two ways it can work.
You just touch the key. This is a non-resident key, that can only be used as a second factor, in addition to giving your username and password.
[1/2]
in reply to Michael Lucas

they are like ssh keys, but for webpages.

So, yes, something "you have". Just like ssh, the keys for webpages are way better than the passwords.

The only problem is that they are harder to backup.

in reply to Michael Lucas

you are missing something. You replace something you know with something google or apple have.

They pinky swear to only present it when they are somewhat sure that it might possibly be you who wants to access a thing.

(Sounds like a shitpost but I'm serious)

reshared this

in reply to florian

@florian With the added "benefit" that if you want to take your passkey-ring and move it to something else that isn't Apple/Goog, well, no. You neither may, nor can.

I'm a little less concerned about passkey integration when it comes to password-managers with an openly-documented format that would let me take them elsewhere.

But Goog/Apple? Not even once.

in reply to florian

forgot to send this the other day, Re: passkeys

Just found it in my gtd inbox.

mastodon.social/@mhoye/1133183…


This is the most important thing you need to know about passkeys: that "Authorizing Party" box in the spec?

That's not you. They're not actually "your" passkeys.


in reply to florian

it's here in the spec: fidoalliance.org/specs/cx/cxp-…

I don't actually read it that negatively, but then I haven't read the spec...

in reply to Michael Lucas

The people involved are promising an updated spec tomorrow, so one hopes we'll know more then
in reply to Michael Lucas

It's a secure authentication scheme, using public key crypto. The real problem is 1) no way to archive and/or transfer auth tokens to a new device, or a different vendor. I believe they will "back up" to the vendor's own cloud, but you can't take your Goog keys and move them to Apple.

2) The software for managing passkeys have various limits on how *many* keys you can store. This is also why, even though YubiCo was involved, you won't find a YubiKey that can store all of your keys.

in reply to Michael Lucas

Interesting thread... :-)
I thought you choose a passkey so that the mean guys don't need to cut off your thumb or head to get access?!
in reply to Michael Lucas

I think it has to do with public/private keys, so unless you can remember your private key: yes.