Work in Progress Session
Ben Chinowsky, Internet2
Public Key Infrastructure (X.509) Library [libpkix]
Steve Hanna, Sun Microsystems
Steve presented libpkix (http://libpkix.sourceforge.net), an extensible C library for building and validating cert paths. They are looking for project participants. Various research questions are involved; one of particular importance is how you limit the amount of effort expended on pathbuilding. The Mozilla developers are very interested in this work.
The Bear Project
Sean Smith, Dartmouth College
Sean discussed Bear (http://www.cs.dartmouth.edu/~sws/abstracts/msmw03.shtml). This work is designed to address the question, "why should you trust computing that happens somewhere else?" For example, why should I trust a Shibboleth attribute authority to be giving my attributes only to the right people? The client wants to only have to provide a cert; the server doesn't want to spend money, and wants easy maintenance and good performance. The IBM 4758 doesn't solve the server problems, as it's expensive and awkward to code for. The Bear project attempts to provide 4758-like functionality on a standard machine equipped with version 1.1b of the TCPA/TCG TPM (e.g., many IBM NetVistas). Bear is now running with OpenCA in the lab; the code is at http://enforcer.sourceforge.net. Smith noted some weaknesses: Bear is probably vulnerable to power analysis; unprotected systems between the client and server will create vulnerabilities; and there are probably holes in the OS code. AEGIS could address some of these weaknesses. A revised and updated Bear paper will appear in ACSAC in December 2004.
Domain Name System Security (DNSSEC) Update
Sam Weiler, SPARTA
Finally, Sam followed up last year's WIP session on DNSSEC with a discussion of what it will take to motivate DNSSEC deployment. Security is expensive to implement, and Weiler pointed out that with security you're basically "buying brittleness" anyway. Three positions were expressed: