[Nsi-wg] The POODLE attack on SSLv3

[Henrik Thostrup Jensen]

Hi

Some of you have probably seen this:
http://googleonlinesecurity.blogspot.com/2014/10/this-poodle-bites-exploiting-ssl-30.html

As we have mandated use of TLS 1.0 (which is the successor to SSLv3), an 
NSI implementation should not be vulnerable.

***
If you are responsible for an NSI implementation, please double check that 
SSLv3 is not allowed (the default contexts often allow this).
***

AFAICT even NSI agents supporting SSLv3 are not vulnerable to the attack 
as we authenticate the client and do not use HTTP session keys (the POODLE 
attack uses single-byte leaking to grab a session key by inserting 
requests into a unencrypted side-channel and reusing it in a new session).

Further, there is some rumor mongering concerning TLS 1.0/1.1 being 
disabled some places. These two have a lot of similarity to SSlv3, but are 
NOT vulnerable to the same attack. While I don't think they can be 
vulnarable to a similar attack (but I am not really qualified to guess), a 
lot of clever people will be looking into creating variants of this attack 
in the next months. So consider supporting TLS 1.2 sooner rather than 
later.


     Best regards, Henrik

  Henrik Thostrup Jensen <htj at nordu.net>
  Software Developer, NORDUnet