September 10, 2003
Attendees
* Jim Jokl, U. Virginia
* Eric Norman, U. Wisconsin
* Steve Worona, Educause
* Shelley Henderson, USC
* Eric D?, Columbia
* Jeff Schiller, MIT
* Mark Franklin, Dartmouth
* Steve Carmody, Brown
* Jeanette Fielden, Internet2
* Renee Frost, Internet2
* Neal McBurnett, Internet2
* Steve Olshansky, Internet2
* Ken Klingenstein, Internet2
Discussion
The Usher CP discussion focuses on the differences between PKI lite and a heavy weight policy. Is USHER is going to put requirements on the campuses or just let the campuses follow the lite model? The CP and practices are about how USHER itself is operated.
Steve delineated what he sees as two use cases. The first is InCommon and simplifying the process of managing the trust fabric. It's a single CA and doesn't have to have anything to do with broader issues in Higher Ed. The second case relates to S/MIME. Although there are not many S/MIME deployments, schools are thinking about it and the implications of inter-campus S/MIME. They are interested in making it easier to manage the trust fabric.
For the federal bridge there a lot of very heavy requirements. Do you impose requirements on everyone in order to serve the two or three people that need access to the federal bridge? Or do most campuses just want a CA to enable interaction with other campuses to enable research collaboration. For grant proposals with the federal government it should be possible to build on top of the PKI infrastructure.
There needs to be some level of confidence that schools can have in the USHER CA and what physical level of security it's run at. The USHER CA needs to be run securely at a level beyond PKI-lite. There needs to be a clear explanation of how you're protecting the private key and tying into the I&A process on the subordinate CAs. The initial level will likely not translate into a federal level of assurance at medium or basic since there are expensive auditing and physical security requirements, which may discourage institutions from participating.
The first USHER CA will follow a CP from the FBCA "Citizen and Commerce" school which is similar to "HEPKI-Lite". It would be a short CP without extensive restrictions on the CPs of the subordinate CAs and detailed legal and technological prescriptions. It will enable institutions to build on the capabilities of InCommon and Shibboleth and establish new sorts of federated trust. Higher Ed will be able to deploy S/MIME and similar inter-institutional technologies and move away from password-based and IP-number-based authorization.
As demand and expertise increases additional CAs will be added that correspond to progressively higher levels of assurance such as the Federal "Basic" or "Medium" LoA using the full HEBCA-compliant HECP.
Eric indicated that it needs to be discussed how the public keys and certificates will be deployed and verified. Jeff is still working on hardware issues to find a device that supports 2048 bit key loading. He's also considering how a safe might be used to run an on-line server in lieu of a secure room. The current focus is getting the Rainbow token to run in house under Linux.
There will need to be documentation that follows a standard format that incorporates a CP, a CPS and standard practices so people can understand what is being offered. It is not clear if it will be a kind of a formal policy and practices that fits a model, or more simply documenting what everyone thinks is a good technical solution and the I&A process as policy practices.
Neal has been trying to
find time to look at the
PKI challenge which came
up with a number of specific
interoperability issues
and if they've been considered
or not.
Jim indicated he could help
with that.
[AI] A formal last call
on Jim's profile posted
on the HEPKI page will be
sent out.
Jim has received some responses
to his e-mail about S/MIME.
If anyone knows of smaller
limited S/MIME projects
they would be welcome as
well. He would like to see
an updating of the S/MIME
client table to include
Pine, Mulberry and any other
e-mail clients. Updating
can be handled via the e-mail
list.
[AI] All. Please let Jim
know of any other schools
doing S/MIME that are not
currently listed.
Jim will co-ordinate some of the audit components for the training documents for Campus CA's and Internal Audit.
Neal asked if anyone has
a better description of
Open SSL and FIPS 140 certification
than is in the press. What's
unclear is how many applications
that people actually use
are actually running in
FIPS mode. What is different
when an application is running
in FIPS mode?
[AI] If anyone has a product
running in FIPS mode please
let Neal know.
There has been some work
done at Dartmouth putting
patches in the LINUX kernel
to turn it into a poor mans
secure co-processor which
might be of help with the
FIPS certification. The
Slashdot entry about Dartmouth's
Linux "Open-Source
Virtual Secure Coprocessor
based on TCPA" is available
at:
http://developers.slashdot.org/developers/03/09/10/0255245.shtml?tid=106&tid=126&tid=172&tid=185.
It has a link to the Dartmouth
paper and to several by
IBM.
Shelley gave an overview
of USC and how they've responded
to the announcement that
Globus is shutdown their
CA. They've created an institutional
CA with a CP based on PKI
Lite meant to have a lifespan
of about six months to allow
them to gain experience.
The CA uses OpenSSL, and
is set up so that only two
people on staff can actually
issue certificates. Emergency
access is allowed if neither
of those two people is available,
but it requires two other
staff members to unlock.
The CA machine is not in
a vault but is non-staff
inaccessible and connected
to the campus network. Initially,
the USC CA will issue certificates
only for hosts and persistent
processes/services, not
for users. User PKI certs
are created using kx.509
with authentication based
on Kerberos.