At 2:46 pm -0800 2000-12-15, L. Sassaman wrote:
On Fri, 15 Dec 2000, Dave Del Torto wrote:
However, if the intent is to "mask" the presence of Bob's key on
the keyserver in lieu of Deleting it, it's hard to see what sort of
keyserver response behaviour would prevent Eve from trying to
determine the presence of Bob's key on the server -- by deducing
that information using the very 0x40 flag you describe in
conjunction with the keyserver behaviour. This simply shifts the
policy focus from the pksd's delete policy to its search policy.
All Eve would need to do is upload the public component to the
keyserver in order to determine the status of the subpacket 23
enable/disable flags: if the key is present but disabled, it will
"vanish," confirming that it's disabled. If not present, it will be
returned by a subsequent search. A rudimentary traffic analysis
technique.
Where is the threat here?
The threat, perhaps a theoretical one at this point in time, is
that if you're trying to mask the very existence of your key on
one or more keyservers (for example to thwart an improperly coercive
demand for keying material from rogue officials), you will not be able
to do so with these flags. You would only be able to do this by
deleting them via an LDAPS (or similarly secure) keyserver connection.
BTW, the trail of evidence that Deletion would involve, reminds me of
another thread I wanted to raise (wearing my CryptoRights hat) with
the keyserver managers, regarding the need for standardized
mechanisms permitting secure, UNauthenticated, UNlogged connections
to keyservers. Deleting a key would still require a passphrase, but
it would permit one to clear a key if one was under threat.
The reason I prefer a "disable" approach to a "delete" approach is
that deleted keys can always reappear. The behavior you describe
seems to me to be the appropriate one.
We can agree on that much. I guess I'm just applying a more stringent
threat model -- speculating that regimes like RIPA may someday soon
negatively impact the human rights of crypto users. I still like your
key-disable flag idea, Len.
dave