[clue-tech] Some thoughts about GnuPG.

[Jed S. Baer]

On Sat, 04 Sep 2010 17:32:31 -0600
David L. Anselmi wrote:

> To be at risk the passphrase has to be cracked *after* the person
> shared their key with an adversary (probably without being aware that
> it needed to be revoked).  So first, let's make OpenPGP widespread.
> Then people can suffer from lack of care with their keys.  Then they
> can learn to take care of the keys.  (Not much different than the
> current problem with identity theft, except that *you* are suffering
> from the lack of care by *others*.)

Okay, now that I understand what you wrote, instead of what I read past
the first time  .... :)

No, the key doesn't have to be shared with an adversary, unless your
definition of being shared includes their machine being lost, stolen, or
confiscated by agents of the government.

So let's say that Alice creates a key-pair using a weak passphrase. I
send Alice my sooper-seekrit stuff -- let's say I've rediscovered the
formulation for Reardon Metal. Industrial spies from Elbonia break into
Alice's office and brute-force the passphrase, thus gaining access to my
revolutionary formula. Oh noes! My business plan is ruined!

Or, we could be a bit more realistic and think about a Chinese dissident
sending messages to an ex-pat reporter in another country.

The identity theft issue comes into play with a 3rd party continuing to
use the key and pretending to be Alice after the theft or confiscation of
her machine. Presumably, in such a scenario, she'll not be in a position
to issue her revocation key.

I also am not too sanguine on the prospects for learning good key
management after the fact. This seems, to me, to mean we're going to use
good crypto, except we can really use it, because we don't really trust
the security of the key. This kinda dovetails into why I dislike the term
"web of trust" and assigning trust levels to keys. What is it we're
trusting? Do we really expect lay persons to grok what is meant by what
is a term of art in the PKI field? Hey, I've read up on this stuff, and
even I don't completely understand what it's accomplishing. If I have
someone's public key on my keyring, and it's signed by DaveA, and I know
that DaveA wouldn't sign a key without physical identification, then I
feel good that that key really is the key for person X. Add even 1 level
of indirection there, and then ... ? But back to the key management
thing, it's easier to learn good habits from the start, than have to
unlearn bad ones later. Then there's the additional barrier of saying,
"Oh, now I have all these messages I can't decrypt because the key's been
revoked." People just aren't going to deal with that inconvenience.

Or am I still missing something?


-- 
Ok, so we should be thinking of a lovable, cuddly, stuffed penguin
sitting down after having gorged itself on herring. Still with me? 
 -- Linus Torvalds