Re: kernel.org status: establishing a PGP web of trust

[Valdis.Kletnieks@vt.edu]

[-- Attachment #1: Type: text/plain, Size: 2895 bytes --]

On Sat, 08 Oct 2011 01:02:13 EDT, Jon Masters said:

> What I'm saying is that unless you sign something (random text, my
> actual key(s)) in my presence, I can't actually know it was you I was
> dealing with or someone else claiming to be you (or your identity).

Now see, this is *exacltly* why security people have to be pedantic about
stuff.  What you originally asked for was "sign random data to demonstrate
control of the key", and I pointed out that being able to sign a key was as
good as being able to sign random data to prove control of the key.

However, it now turns out that your *actual* worry is "binding the person
who controls the key with the person who controls the e-mail", which is
in fact a valid concern for some situations, and as pointed out, requires
a bit more effort to establish.

However, you have to remember that *all* real-life identity proofs are to
some degree probabalistic.  How do you know your lawyer is a *real* lawyer
and not somebody with a fake degree?  Mostly because the other lawers
in town and the judges are convinced he's a lawyer too.  Same goes for
your doctor - how do you know he *really* went to med school?  Even calling
the med school and verifying only proves that somebody with that name went
there that year.  And yes, every year we hear about a few fake lawyers and
doctors on the news.  But society seems to muddle along just fine anyhow.

(Incidentally, something most people don't realize is that the entire debit/credit
card industry is *designed* with the assumption that between 2 and 4
percent of all transactions will turn out to be fraudulent in some way.  So
everybody keeps a small buffer for chargebacks and life goes on).

Similarly for PGP -  it's not *that* hard to create a totally new e-mail
account, a new name, get a fake ID, go to a few key signings, and have a nice
validated bogus PGP ID.  However, do we really *care* in that case?  Probably
not.  What we *care* more about is somebody creating a fake ID in somebody
else's name.  And that turns out to be rather self-limiting - all it takes is
*one* person to send the real person an e-mail that says "Glad to meet you at
the keysigining last week", and the ruse is revealed when the real person
realizes he wasn't *at* the keysigning.

And even if the identities *are* perfectly confirmed, that doesn't remove all
the risk.  You can have somebody you've known for years, do background checks,
and be 100% convinced he's the real person.  But if you then use PGP to encrypt
the secret plans for the revolution and e-mail it to him, PGP says nothing at
all about whether he's in reality a government mole who's infiltrated your
orgainization...

That's why nobody worries too much about the "is it really him?" side - the
world is full enough of properly identified but duplicious people to worry
about the few who have fake identities and are duplicious. ;)


[-- Attachment #2: Type: application/pgp-signature, Size: 227 bytes --]