Security defaults

[Brent Putman]

On 5/18/14 3:03 AM, Cantor, Scott wrote:
> Yeah, it's just the primary purpose of blacklisting it is decryption, to
> prevent the IdP from compromising its own key by attempting a decrypt with
> that algorithm. Blacklisting on the signing side is actually more of a
> good citizen kind of thing, to avoid hurting others.

Ok, I did move RSA PKCS 1.5 into blacklisted status for both encryption
and decryption.


>
> The SP doesn't have independent lists for all  the modes, it's just a big
> list of identifiers for all algorithms.


I thought about that, but due to the way the signing and encryption
resolvers work, I decided to keep all 4 categories separate.  That's to
be flexible and account for the case where maybe you're ok with signing
to someone with SHA-1, if that's all they support, but you're not ok
with accepting/validating signatures with SHA-1.  Of course, in the
Spring config, one could easily just define one common <util:set> and
inject that into all 4 places, so that would effectively be the same
thing as a unified list.


While we're on the topic:  I tried to find in the wiki some definitive
docs for the SP on: 1) what algorithms are blacklist be default 2) what
algorithms are effectively supported by various versions of the SP.  But
I didn't see anything.  Do they exist and can you point me?  For #2 I
did see some dicussion about what is supported by various RHEL OpenSSL
versions, but that's only relevant if the platform lib is in use - don't
we ship our own?  Or is that only on Windows?

For OSJ-77, I was doing some testing around supported algorithms in
Java.  The quick summary is:  if you want support for AES-GCM and/or RSA
OAEP 1.1, you need either Java 8 or the Bouncy Castle provider. 

With OSJ-77 (aside from some more testing), I think we're pretty good
now on not accidentally trying to use an algorithm that the runtime
environment doesn't support.  For example, if some SP's
KeyDescriptor/EncryptionMethod has AES-GCM or RSA OAEP 1.1 listed, it
won't get used unless the JVM environment actually supports.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://shibboleth.net/pipermail/dev/attachments/20140520/f333adbe/attachment-0001.html