OT: crypto, auth, CAs, web-of-trust, and phony certs

["Dr. Ed Morbius" <dredmorbius@gmail.com>]

Apropos some of the recent discussion we've had here of various Debian
signing keys.

A major CA (certificate authority) has issued fake SSL certs for
Google.com, Yahoo.com, and Skype.com (and apparently 6 other sites)
after its signing keys were compromised.

    http://threatpost.com/en_us/blogs/phony-web-certificates-issued-google-yahoo-skype-others-032311

Under the CA / SSL model, you trust a website because you trust the CA.

What will likely happen now is that these keys will be specifically
noted as compromised, though this will have to be handled through
various SSL/TLS libraries and client applications.

A somewhat analagous situation occured with the Debian project when an
entropy-limiting bug introduced into SSH key generation resulted in a
very low number of unique keys being generated:

    http://jblevins.org/log/ssh-vulnkey

The generated keys can now be checked for with the 'ssh-vulnkey'
command, and ssh client and server software are largely now configured
to disallow connections based on blacklisted keys (either server or
client keys may have been compromised).

Which is to say:  crytpography is important, but it's not the only
attack vector.  And good crypto, both in its implementation and
execution, is hard.

-- 
Dr. Ed Morbius, Chief Scientist /            |
  Robot Wrangler / Staff Psychologist        | When you seek unlimited power
Krell Power Systems Unlimited                |                  Go to Krell!