OpenSSL heartbleed bug / Shibboleth implications
Cantor, Scott
cantor.2 at osu.edu
Mon Apr 7 23:25:11 EDT 2014
A very serious bug in OpenSSL 1.0.1 was announced this afternoon:
http://heartbleed.com/
The actual direct impact on the Shibboleth software packages is relatively minimal in comparison with the fallout from this. The only distribution of OpenSSL included with any Shibboleth software is the Windows SP.
I'm not waiting to produce an actual advisory on this before saying something because this is a major, major bug and it's public.
I am working to prepare a patch for this (I had no advance warning) and it will be done as soon as I can produce it. It will *only* apply to the supported SP version, which is 2.5.3. Anything older than 2.5.0 didn't include an affected OpenSSL version, but any 2.5.x version will need to be updated to 2.5.3 and then patched.
Any other SP version is still vulnerable if used with OpenSSL 1.0.1, but I don't control the process of obtaining an update, so that will depend on your OS or local build.
On the IdP side, this is really a matter for deployment considerations. We don't provide the actual web server and TLS implementation for the IdP, so you would need to evaluate your choices there and determine whether OpenSSL 1.0.1 is implicated. Obviously pure Java solutions hosting the IdP are not, though some Java containers can be configured to use OpenSSL as a TLS stack for performance reasons.
As to the implications, this is a very severe bug, and has apparently been shown to leak the private key used on the server or client. In the case of an IdP, that usually means the potential exposure of *the* signing key because that key usually doubles as a server key for SOAP traffic on a second port.
In the case of the SP, there is, I think, somewhat less risk because the SP doesn't generally contact arbitrary servers that might be used to attack it, but that doesn't guarantee safety.
In the security parlance, keys at risk are basically considered compromised and the official advice would have to be to revoke and replace them. I would imagine that federations will be moving on this to help people understand and react to this, but I felt an obligation to say something in the interim, given the gravity of the bug.
-- Scott Cantor
Shibboleth Project/Consortium
More information about the announce
mailing list