From: "Frank O'Dwyer" <fod@brd.ie>
Date: Sun, 7 Jan 1996 08:55:08 +0800
To: "tcmay@got.net>
Subject: RE: Revoking Old Lost Keys
Message-ID: <01BADC96.AA0A0B20@dialup-169.dublin.iol.ie>
MIME-Version: 1.0
Content-Type: text/plain


On Saturday, January 06, 1996 07:19, Timothy C. May[SMTP:tcmay@got.net] wrote:
>At 9:47 AM 1/6/96, Frank O'Dwyer wrote:
>>On Saturday, January 06, 1996 09:18, Timothy C. May[SMTP:tcmay@got.net] wrote:
>
>>>Basically, you are screwed. Any revocation you attempt will not be trusted,
>>>as we will suspect the new "you" to be an attacker, perhaps an agent of the
>>>NSA or the Illuminati. In the view that "you are your key," the old you no
>>>longer exists.
>>
>>This is true, but the "old you" can be resurrected if you can get enough
>>people to believe your new key using any out-of-band means available
>>to you.  You can also put a comment in your new key's uid explaining the
>
>Could you explain how "enough people" can get around a basic
>feature/limitation of the current PGP web of trust? Who, besides the
>originator, can revoke an old key? How many does it take?

I wasn't referring to revoking the old key, but to introducing a new one
and letting the old one fall into disuse. I think this can sometimes be 
done even if you've lost access to the old key, albeit in a painful out-of-band 
fashion.  It does depend on the application, though, and if the (relevant 
portion of the) web of trust was very large, you might find that the old key 
kept popping up and you kept getting mail (or whatever-it-is-you're-encrypting) 
that you couldn't read. (and/or some people wouldn't believe your signatures).

Basically, PGP's revocation model is broken unless you create a revocation
cert. at the time you make your key, and keep it safely somewhere in case
you need it.  Even then, as time goes by the keyring keeps accumulating all 
these extra packets and growing without bound.  It's not just PGP--all long-lived 
certificates are hard to revoke (for example X.509's revocation is also clunky).  
It's just that PGP's certificates are particularly long-lived, and PGP's revocation is 
particularly broken. Luckily the data formats do allow for a validity time, and a 
revocation of a key's countersignature, so this can perhaps be fixed sometime.

>If a bunch of the "alleged" friends of Bruce could do this, could they not
>revoke the key of someone they simply wish to hassle?

Well, see above. The key is not revoked.  But a bunch of people _could_
attempt to introduce a key under the name of someone they just wanted to 
hassle.  The conspiracy doesn't have to be especially large. For example, it 
would be easy for me to invent a key for you and have _my_ friends believe it
even in spite of your real key being on their keyrings.  It wouldn't be so easy 
for me to get _your_ friends believe it.  In Bruce's case, he'd be trying to do
a similar thing, except that the key'd really be his, and more people'd be
likely to believe him (especially his friends, and their friends, and so on).

>I agree that a new key can be generated, and a new "Please use this key,
>not the other one" message sent, and this may work, but I don't believe
>this revokes the old key and removes it from the keyservers. I could be
>wrong, as I am certainly no expert on the keyservers.

I think you're right.  The key will still be out there.

>The question is: is there a "majority vote" mode on the keyservers that
>causes them to remove a key if enough people claim it is no longer valid?

I don't think so.  At best, you might be able to convince the admins to
manually delete the old key from the server's rings (assuming the software is able
to do this).  Even then, the key might keep popping back up, for example
if you had countersigned other people's keys with your old key and they kept
uploading their key with additional signatures.  A practical solution
might be for the key servers to automatically remove keys older than X 
years (or some time limit related to the key size).  Ultimately though, what 
is needed is a new revocation model (maybe implementing the unused fields 
in the PGP certs is good enough to begin with).

Cheers,Frank O'Dwyer
fod@brd.ie                          http://www.iol.ie/~fod