Icecast dev - Nov 2014 - Proposed openSSL usage improvements

Hi everyone,

Prompted by the fact that addressing some of the recent SSL problems
actually would benefit from also changing things on how openSSL is used
(not just updating the library), I started looking into some improvements.

The tracking ticket is:
https://trac.xiph.org/ticket/2070

To sum it up:
 - hard disable SSLv3
 - hard disable compression
 - new default cipher list
 - enable forward secrecy

This should basically address anything that can be done to further
alleviate recent attacks. If you know of something else, please let us know.

As this is security relevant code, I'd like to solicit more eyes on the
patches.

Side note: I haven't verified it, but I suspect that with the patches
applied Icecast will no longer build against openSSL 0.9.x. I personally
find this acceptable, that version is pretty much on its way out and
e.g. in the Debian context only available in old-stable aka Squeeze.

This will largely be part of the 2.4.1 release that we're currently
preparing. (PFS might not make it)

Cheers

Thomas

On 11/02/2014 05:56 PM, "Thomas B. R?cker"
wrote:
> Hi everyone,
>
> Prompted by the fact that addressing some of the recent SSL problems
> actually would benefit from also changing things on how openSSL is used
> (not just updating the library), I started looking into some improvements.
>
> The tracking ticket is:
> https://trac.xiph.org/ticket/2070
>
> To sum it up:
>  - hard disable SSLv3
>  - hard disable compression
Landed ready to be released in 2.4.1.

>  - new default cipher list
Went with
https://wiki.mozilla.org/Security/Server_Side_TLS#Intermediate_compatibility_.28default.29
in the end.

Previously planned using this:
https://hynek.me/articles/hardening-your-web-servers-ssl-ciphers/#fnref2

Testing against Qualys gives me identical results for both.
We might upgrade to the "Modern" Mozilla string in the future, but as
of
now that completely breaks our HTTPS functionality. I suspect, because
we don't properly support all elliptic curve ciphers yet, which is on
our to do list though.

Thanks to Basil Mohamed Gohar for pointing me towards the Mozilla cipher
lists.


Cheers

Thomas

PS: 2.4.1 is fix/feature complete and we're now working on documentation
and testing. Release should happen within days.

On 2014-11-09 12:37, "Thomas B. R?cker" wrote:
> On 11/02/2014 05:56 PM, "Thomas B. R?cker" wrote:
>>   - hard disable compression
I assume this is only header compression (where the vulnerability was) 
and not content compression?


-- 
Roger "Rescator" H?gensen.
Freelancer - http://www.EmSai.net/