PaulDotCom mailing list archives

SSL certification chain caching


From: jim.halfpenny at gmail.com (Jim Halfpenny)
Date: Mon, 19 Apr 2010 21:00:32 +0100

Hi all,
Given all of the talk round SSL, root certs, MitM attacks I got to
thinking about how to mitigate this. I read the paper by Christopher
Soghoian and Sid Stamm (http://files.cloudprivacy.net/ssl-mitm.pdf)
and there are some obvious flaws in the Trust On First Use concept,
particularly if you are on a high risk network or in a high risk
country with draconian Internet censorship like Australia. So why not
compile a list of the SSL certificates in the wild and the chain of CA
certificates? If a certificate, the CA chain or other metadata differ
significantly from the cached version raise a flag or score the
likelyhood of it being compromised/altered/Chinese.

+10 for a new cert
+50 for a new CA cert in the chain
+666 for signed by a nation state CA

Yes, the major fatal flaw is again trust; do you want to start
depending on a 3rd party to vet your SSL certificates? The key factor
is not depending solely on the CA to verify authenticity; querying
certificate repositories (think http://pgp.mit.edu) in addition to
checking the SSL certificate could offer another layer of defense.
Offline databases could be disseminated out of band so end users could
examine their SSL connections for MitM tampering in hostile
environments. I've lots of ideas and am working on some of them at the
moment, any suggesions, advice and help welcome!

Regards,
Jim


Current thread: