[7973] in Kerberos
Re: Kerberos 5 and S/Key
daemon@ATHENA.MIT.EDU (Dan Geer)
Mon Sep 16 09:05:43 1996
To: joek@cybersafe.com, mas@ucla.edu
Cc: kerberos@MIT.EDU
In-Reply-To: Your message of "Sun, 15 Sep 1996 19:02:43 GMT."
<51hjoh$1m7@kerby.cybersafe.com>
Date: Mon, 16 Sep 1996 08:40:51 -0400
From: Dan Geer <geer@OpenMarket.com>
Joe Kovara (joek@CyberSafe.com) writes:
mas@ucla.edu (Michael Stein) in comp.protocols.kerberos wrote:
> What's the point? My understanding of Kerberos is that anyone can
> request a TGT from the kerberos security server, but the reply is
> going to be encrypted in the users secret key (his password).
Reusable passwords are less secure than single-use/one-time
passwords because reusable passwords are subject to many
relatively simple attacks. The password itself is the object
of the attack--which obviously results in compromising whatever
the password is protecting--a seemingly academic, but important,
distinction.
To put it another way: static targets are easier to hit than moving ones.
Well said, Joe.
If I could add for this Michael's edification; the single
most important thing to prevent "relatively simple attacks"
is good password choice. (Relatively simple is a relatiive
term -- it is relatively simple for me to powder a large
boulder with a sledgehammer as there are no intellectual
problems lurking there, just work, work and more work.)
Michael, if you use full strength Kerberos with good
password quality control and you'll be so much better than
an unimproved site that bandits will go elsewhere.
Good password quality includes:
no dictionary words (probably means nothing pronounceable)
no local acronyms (for joe, "OCSG" would be a bad choice)
minimum length (like 8 characters)
minimum variation (like must have alphas, numerics and punctuation)
maximum lifetime (aging, as it is generally known)
minimum reuse cycle length (cannot use same password often)
The use-once password idea does moot some of the quality
control issue because it substitutes sharp containment-in-time
for the other limitations listed above. When operating
in a setting with a high probability of eavesdropping,
such as a conference terminal room where security was not
a design criterion, use-once credentials may be uniquely
valuable.
Use-once passwords do have a higher management cost,
however, as the constantly changing password requires good
synchronization between server and client. This
synchronization can be expensive, such as when needing to
re-synchronize on demand (I lost my S/Key cribsheet, or
my SecurID's battery died, etc.).
I might add that key distribution centers like Kerberos
do have an advantage here over public key systems in that
the PK systems really don't have a prayer of enforcing
password quality control since there the password protects
the private key half, something central security services
never see or hold.
--dan
