3

I find the common procedure of key signing parties a little awkward:

Person A presents the passport and a fingerprint "This is my key!". Person B checks the passports and decides a level of trust to sign the key at home.

Now the only thing person B can really confirm is that person A was present (based on trusting the passport) claiming that the printed key is his/hers. The only mapping to the key is by name though - which might not even be unique.

In fact person A could have looked up someone that has the same name and payed him/her to go to that key signing instead. Unless the key has a photo id (very few have) no one would notice. Maybe not very likely or even an attack vector but an example to show the problem.

The only way to confirm that person A has access to the private key is for her/him to sign something. Yet signing (or even just bringing keys) is usually not permitted/wanted on most key signing events.

But wouldn't that make much more sense? No trust delegation to a passport. No visual matching to a photo. What am I missing?

Would you consider the common procedure more secure than signing on the spot? Can you imagine a better alternative to the common procedure?

5
  • "...more secure than signing on the spot?" You already mentioned that the participants won't have their private keys with them, so signing on the spot isn't an option. In theory the users could wait until they go home, then wait for a proof message before signing another person's keys.
    – Mr. Llama
    Jul 11, 2018 at 22:05
  • Well, my point is that this restriction might not be very useful. So it could be made an option by bringing the keys.
    – tcurdt
    Jul 11, 2018 at 22:15
  • I don't see how sending a message later is proof.
    – tcurdt
    Jul 11, 2018 at 22:17
  • The point is to verify the identity of the person and match that identity to the claim that they are related to the public key. There would be no advantage for an authenticated individual to misrepresent their key.
    – nbering
    Jul 12, 2018 at 1:01
  • There is no way to proof that claim by just looking at a passport. The fact that the person in front of me has in fact access to that key is purely based on matching a name. I don't see how to turn this into an attack vector - but that doesn't change the fact there is no real proof.
    – tcurdt
    Jul 12, 2018 at 8:08

1 Answer 1

3

Key signing parties are, indeed, kind of useless if it's between people who have never encountered each-other before. Unless you're an expert at recognizing forged documents, you would never be able to tell a real passport from a well-done fake passport (or, worse, a real laminated ID from a fake laminated ID, if it's from another state of province). Also, it's in the human nature to act agreeable when dealing with other humans, so nobody is going to poke at an ID trying to see if it's a fake or a real thing. The shut-in human nerd you'll be dealing with will just smile, say something like "yeah, it looks like an ID" and agree sign your key -- all while feeling super awkward and embarrassed.

The OpenPGP standard actually has a mechanism to indicate how much effort you've put into verifying someone's identity. If you run gpg with --ask-cert-level when signing a key, it will offer you a way to indicate how much identity validation you've done, from "none at all" to "very extensive verification." This option is off by default because extremely few people use it correctly, so it's pretty meaningless these days.

Meaningful keysigning parties do not involve passports or any other kind of identification documents, because they happen between people who know each other from previous engagements. For example, when someone like Linus Torvalds signs a key belonging to another Linux kernel developer, he does not do it to indicate that "this is Dirk Hohndel because I checked his government-issued ID," but to say "this key belongs to a person whom I know as Dirk Hohndel." It is not important to other Linux kernel developers under which name Dirk files his taxes -- all they really want to know is that "this is the same Dirk who works with Linus Torvalds."

So, if you're attending a keysigning party where nobody knows each-other, just leave -- any key signatures obtained at such events will have very little value.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .