Transferring identity to a new public key

[David Shaw]

On Mon, Feb 16, 2009 at 12:10:32PM +0100, Jonas Islander wrote:
> When you suspect your private key may be compromised, it's obvious
> that you should revoke the key pair, upload your revocation to the key
> servers, and generate a new pair. But what is "best practice" for
> telling people about your new public key - transferring your identity
> to it, so to speak?
> 
> Is there any point in adding a self-signed ID saying "Key compromised
> - please use key with fingerprint xxxxxxxxx instead" before revoking?
> 
> I'm thinking it's pointless, since an attacker could do the same, and
> use it to transfer someone's identity to a new public key, which the
> rightful owner cannot revoke.

Yes and no.  Such a message is okay so long as the person seeing it
treats it as a hint - that is, to go and fetch the new key, and then
build a brand new trust path to this new key.  As you note above, it
is pointless to assume the new key is good just because the old key
tells you.

> Am I right in thinking that anyone seeing a user ID of the form
> "Please use key with fingerprint xxxxxxxxx instead" should ignore it
> (since it may be an attempt to permanently steal someone's identity)?

They should ignore it (or more likely try and contact the keyholder
and figure out what is going on) if they cannot build a valid trust
path to the new key that does not go through the old key.

> Am I right in thinking that someone whose key may be compromised,
> should simply revoke it and start over from scratch with a new key
> pair, proving their identity to each and every person signing it?

Yes.  It is bad practice to sign a key just because they signed a
previous key owned by the same person.  You should check each time.

David