access to storage keys, NOT comms keys! (Re: What's really in PGP 5.5?)

[Adam Back <aba@dcs.ex.ac.uk>]

[made a few snips to the CC line, still Cc of cypherpunks]

William Geiger <whgiii@invweb.net> writes:
>    at 11, Adam Back <aba@dcs.ex.ac.uk> said:
> >1. to allow access to important material lost in the mail system in the
> >event that an employee is hit by a bus
> 
> >Argument 1 seems pretty flimsy to me.  I reiterate my comment in an
> >earlier post: who in their right mind keeps their _only_ copy of ultra
> >valuable company information bouncing around in the email system?  Did
> >those arguing for this position not notice that sometimes email gets lost
> >in transit?
> 
> Well lets take the flip side of this: Who in their right mind encrypts
> ultra valuable company information and then leaves the plain text on their
> computer?? 

Lots of people.  See, what goes over a public network in the clear is
much more vulnerable than what sits on your disk.  I encrypt
communicated copies of things which aren't encrypted on the disk
myself.  I suspect you do too.

But, more to the point, my argument was that keys should have
segregated uses.  One key for storage, another for receiving encrypted
emails.  I wasn't saying that you wouldn't encrypt your archived sent
& received email.  I _was_ arguing that a better way to archive email
securely is to encrypt it with a separate storage key.

> I have an outbox full of encrypted messages that are encrypted
> to both the recipient and to my key (Encrypt-To-Self Option). 

Bad move dude.

See you might have a 100 bit entropy passphrase, but the recipient
might have a password of "fred".  You've conveniently archived all
your email, and you've left it decryptable by a hodge podge of other
people with unknown level of care about your level security.

Say perhaps fred deleted the email after reading, even though he has a
poor passphrase.  You have just screwed your own security.

Similar problem if it is you that has a passphrase of "william".  I
won't thank you when the feds decrypt your email to me, thanks to a
you having a poor passphrase.  (Not that I'm suggesting you do).

If your email archives are encrypted with storage keys, you avoid all
these problems, and avoid GAK arguments at the same time.

> If you are going through the trouble of encryption why would you
> want to leave plain text lying around??? One needs to remember that
> e-mail is not just communication but communication *and* storage.

Nope.  Email is communication.  Archived email is storage.  Use
communication keys for communicated email, and storage keys for
encrypting archived email.

This is a very important point, and I can't fathom why so many people
who are otherwise on the ball are not getting it.  If you don't escrow
any communication keys, but do escrow storage keys, the GAKkers don't
get what they want, and you get all the functionality you need.  They
actually have to break into premises, and take disks, and supoena
keys.

Right?  Simple enough isn't it?

> >2. to allow management to spot check the emails being sent and received
> >
> >A less GAK friendly way to implement it, and a more secure way would
> >be to archive for a while the session keys. The security advantage
> >being that the email doesn't go out with the session key encrypted
> >to 2 long term public key encryption keys.
>
> I have seen no evidence that encrypting to multiple recipients is any less
> secure than encrypting to one. 

Of course it's less secure.  It's less secure almost by definition.

Lets say you have your communications encrypted with only your key,
and there is a small probability call it p1 that your key is
compromised (key board sniffer virus, hidden video cam, typing
passphrase whilst on phone (yes?), whatever). 

Well if you encrypt to another key, say a corporate escrow key, there
is an additional chance, call it probability p2, that your security
can be blown by the corporate key being compromised.  So long as the
p2 is greater than 0, which I'm sure you'll agree it is, however
small, then you have less security by using multiple encryption.

> If there are serious security implications in doing so then it
> affects *all* versions of PGP and not just 5.5.  I find it odd that
> this issue is only now being brought up with 5.5 and never mentioned
> with previous versions.

I've been arguing against using encrypt-to-self for ages.  It simply
makes me cringe when people send me email which is encrypt to self.

> One thing I would like to see added to this set-up is secret sharing of
> the corporate private key. [details elided]

Sounds like a good idea.

Adam
-- 
Now officially an EAR violation...
Have *you* exported RSA today? --> http://www.dcs.ex.ac.uk/~aba/rsa/

print pack"C*",split/\D+/,`echo "16iII*o\U@{$/=$z;[(pop,pop,unpack"H*",<>
)]}\EsMsKsN0[lN*1lK[d2%Sa2/d0<X+d*lMLa^*lN%0]dsXx++lMlN/dsM0<J]dsJxp"|dc`