define ecdhe/ecdsa aes 128/256 cbc/sha mode as per rfc 4492

[jbremer]

As per commit message it would be great to include this patch and preferably even a new alpha release. As I'm using tlslite-ng as part of httpreplay which in turn is used in Cuckoo Sandbox - where we see quite some analyses hitting this exact issue - it would be great if we can start using a new version of tlslite-ng without having to monkey patch our tlslite-ng==0.6.0a3 setup.

Thanks,
Jurriaan

---

This change is

[tomato42]

Sorry for the long delay, DROWN kept me busy.

In general I have no problem with adding them, some minor issues:

• I think they should be clearly marked as not supported (and a reason for the lack of support)
• I don't see why not add in similar way the rest of ECDHE-ECDSA ciphers, especially the AES-128-GCM-SHA256 ones, especially if decrypting communication with google services is the goal
• Generally the ciphers in the lists are added in the strength order, so they definitely should be put before the NULL or anon ciphers :)
• I have workarounded the python 2.6 issue, so please rebase on current master

[jbremer]

Hi, thanks for the elaborate answer, and no need to worry about delays :)
I rebased the PR for you. I'm all for adding some more constants as well, the following reference could help in that matter https://tls.mbed.org/supported-ssl-ciphersuites.
Regarding documentation I'm not sure what the best way to do that would be, so I'll leave that up to you for now ;-)
Fair enough @ null/anon cipher suites..

[tomato42]

The official list is here: https://www.iana.org/assignments/tls-parameters/tls-parameters.xhtml

regarding documentation: a simple

# not implemented, ECDSA certificate support missing

would be sufficient above the definition of every cipher missing full support (just the name and ietfName entry, the arrays are good as is)

[tomato42]

any chance for update?

[jbremer]

Unfortunately I didn't put any more effort into this issue since the PR, provided it works as expected on my side :-)

[tomato42]

if you want to manage out-of-tree patches, I won't stop you ;)

[jbremer]

You make it sound like the most evil thing, but you're right :-(
Maybe in a couple of weeks from now - quite swamped as well. Sorry for that :)

[tomato42]

ok, I'll remind you later if nobody picks up the ecdsa in general

[tomato42]

In the end I needed it myself, take a look at #151.

Closing as #151 is a superset of this PR.

[jbremer]

Looks good! Thanks :-) Your PR might also explain another issue to me, namely one with decrypting RC4 related TLS streams (namely that, as you stated, those ciphersuites are not supported). Will have to look into that at some point as well..

[tomato42]

RC4 is supported, just not by default. But yes, at least one RC4 ciphersuite that should work was missing...

[jbremer]

Ah, right. I'm sorry to ask this here, but what's required to be done to support RC4? With some quick grep's I see there's OpenSSL-based RC4 support as well as a Python version (which is probably slower but I'm fine with that). Thanks!

[tomato42]

it's caused by this configuration: https://github.com/tomato42/tlslite-ng/blob/master/tlslite/handshakesettings.py#L20

so to change it, you need to add "rc4" to the HandshakeSettings.cipherNames array

[jbremer]

Ah, I see, thanks for the information!