doc/conferences/seminar/outline



outline for 1 hr seminar talk to CS/security academics

 - key-based authentication is here to stay. (e.g. https, ssh).
  - host vs. user

 - raises key management/distribution issues

 - what PKIs are available?  X.509, OpenPGP, SPKI

 - social vulnerabilities - single-signer vs. multi-signer

 - protocol vulnerabilities - single cert vs. multi-cert (server
   vs. client again)

 - utility for group-internal work, phased approach to public


Stream-based communications over the public network have an
authentication problem.  Most data streams are not authenticated in
either direction, and most of those that are authenticated in at least
one direction use authentication regimes which suffer from a range of
known structural problems.

Public-key-based authentication offers security advantages over
shared-secret approaches, but it introduces additional questions of
key distribution, binding, and revocation.  Two common solutions to
these problems on today's network are X.509 certificates (used by TLS
connections like HTTPS) and so-called "key continuity management"
(KCM) (used by popular SSH implementations and the "security
exceptions" interface for some web browsers).  Both of these schemes
present security concerns of their own: KCM has trouble with initial
contact, key revocation, and re-keying; and X.509's single-issuer
certificate format has a systemic bias that selects for unaccountable
third-party authorities.  New work ("the Monkeysphere") extends the
OpenPGP Web of Trust into authenticating stream-based communications
(instead of its traditional message-based environment of e-mails and
files) by means of a protocol-independent overlay.  As a simple,
alternative PKI, the Monkeysphere resolves these failings, and also
provides features currently only available as protocol extensions
(such as SNI).