oss-sec mailing list archives
OpenPGP certifications are identity assertions [was: Re: upstream source code authenticity checking]
From: Daniel Kahn Gillmor <dkg () fifthhorseman net>
Date: Thu, 02 May 2013 12:42:04 -0400
On 05/02/2013 11:24 AM, Alistair Crooks wrote:
Oh, I'm not muddled, I was using the word "trust" as it appears (34 times) in RFC 4880 - mostly relating to 5.2.3.13 "Trust Signature" and 5.10 "Trust Packet".
most OpenPGP signatures in the current web of trust are *not* trust signatures, and trust packets aren't emitted or transferred publicly (they're private indicators used for local keyring storage). So what you're seeing from the public keyservers doesn't have any "trust" information in these senses at all. It is just identity assertions.
And if you seriously think someone who searches for my public key on a webserver, or through mail, or business card, etc, downloads my public key from one of the servers, imports it into their own pubring, signs it with their own private key, then mails it to me, or uploads it to one of the key servers, all without trusting me in any way, then I'll show you a pretty awful stalker (and fairly inefficient one, due to the need to sign my pubkey), a fan boy (which is hardly likely to happen in my case), or someone who is rather sad. (I'm discounting impaired judgement due to the baroque processes involved here, sorry xkcd). i.e. no-one goes to that kind of trouble just to say "I know this person" - that's what facebook and google+ are for.
Hm, i'm not sure which category i fall into here, but i definitely make my OpenPGP signatures as identity assertions ("i believe this key belongs to the person named in this User ID"), and nothing more. I believe that's what other people should be doing too. If you see that i've signed someone else's key, please do *not* assume that this says anything about what i think about their character, their politics, their technical skill, their taste in food or clothing, their financial acumen, or anything else. If you try to infer anything more complex than identity from an OpenPGP certification (i'll call these "identity++" assertions), a number of problems arise: 0) it becomes much more difficult to decide to make (and to decide when to revoke) an "identity++" OpenPGP certification, since your belief about someone's technical skill or competence as a babysitter might change much more frequently (and with more nuance) than your belief in their named identity; as a result, people will make fewer identity++ certifications, which will make the entire certification network less useful in identifying peers over an untrusted network. 1) Some people might be making "identity++" assertions based on financial acumen (they only care about money) and other people might be making them based on culinary expertise (they only care about food), so the system becomes much more difficult to reason about, since each certification has wildly different semantics. 2) With "identity++" assertions, the system starts to leak significantly more detailed information related to the social graph, other than just "i have confirmed the identity of this person". This is reckless and worrisome from the perspective of global surveillance (though it's clear that most people don't care much about global surveillance, given that most facebook and g+ data contains significantly more detailed information than even an "identity++" assertion, so maybe this aspect is largely irrelevant). Please just use OpenPGP certifications as identity assertions, not "identity++" assertions. Anyone who wants to say "this person makes a truly delicious quiche" can say so in some other medium (e.g. g+ or facebook or signed e-mail or whatever). Regards, --dkg
Attachment:
signature.asc
Description: OpenPGP digital signature
Current thread:
- Re: upstream source code authenticity checking, (continued)
- Re: upstream source code authenticity checking Kurt Seifried (Apr 25)
- Re: upstream source code authenticity checking Dag-Erling Smørgrav (Apr 26)
- Re: upstream source code authenticity checking Kurt Seifried (Apr 26)
- Re: upstream source code authenticity checking Dag-Erling Smørgrav (Apr 26)
- Re: upstream source code authenticity checking Alistair Crooks (Apr 26)
- Re: upstream source code authenticity checking Kurt Seifried (Apr 26)
- Re: upstream source code authenticity checking Eric H. Christensen (Apr 29)
- Re: upstream source code authenticity checking Daniel Kahn Gillmor (Apr 30)
- Re: upstream source code authenticity checking Robbie MacKay (May 01)
- Re: upstream source code authenticity checking Alistair Crooks (May 02)
- OpenPGP certifications are identity assertions [was: Re: upstream source code authenticity checking] Daniel Kahn Gillmor (May 02)
- Re: OpenPGP certifications are identity assertions [was: Re: upstream source code authenticity checking] Simon McVittie (May 02)
- Re: upstream source code authenticity checking Kurt Seifried (May 02)
- Re: upstream source code authenticity checking Russ Allbery (May 02)
- Re: upstream source code authenticity checking Alan Coopersmith (May 02)
- Re: upstream source code authenticity checking Russ Allbery (May 02)
- Re: upstream source code authenticity checking Josh Bressers (Apr 25)
- Re: upstream source code authenticity checking Alistair Crooks (Apr 25)
- Re: upstream source code authenticity checking Marcus Meissner (Apr 26)
- Re: upstream source code authenticity checking nicolas vigier (Apr 25)
- Re: upstream source code authenticity checking Alistair Crooks (Apr 25)