Originally posted on securitypitfalls:
After nearly half a year of work, the Internet Engineering Task Force (IETF) Request for Comments (RFC) 7465 is published.
What it does in a nutshell is disallows use of any kind of RC4 ciphersuites. In effect making all servers or clients that use it non standard compliant.
I was passed an interesting article, this morning, regarding hardening secure shell (SSH) against poor crypto that can be a victim of cracking by the NSA and other entities. The article is well written and discusses why the changes are necessary in light of recent Snowden file releases.
I’ve noticed a few of my favorite websites failing with some odd error from Firefox.
The Firefox error message is a bit misleading. It actually has nothing to do with the website supporting SSL 3.0 but the advanced info is spot on. The error “ssl_error_no_cypher_overlap” means that the client didn’t offer any ciphers that the server also supports. Generally when I see this I assume that the server has been setup poorly and only supports unsafe ciphers. In this case the website only supports the RC4 cipher. I wondered why I was starting to see a reversal of removing RC4 from so many websites recently (especially since RC4 is very weak and is on the way out). Apparently these websites all use the F5 load balancer that had a bad implementation of the TLS 1.0 standard causing a POODLE-like vulnerability.
Stepping back for a moment, back in October the POODLE vulnerability hit the streets and a mass exodus from SSL 3.0 happened around the world. I was happy to see so many people running away from the broken cryptographic protocol and very happy to see the big push to implementing the latest version of TLS, TLS 1.2. So with SSL 3.0 out of the way and the POODLE vulnerability being squelched why are we seeing problems in TLS 1.0 now?
Well, simply put, F5 load balancers don’t implement TLS 1.0 correctly. The problem with SSL 3.0 is that the padding format isn’t checked. Apparently in the F5 devices it’s still a problem in TLS 1.0. And while the company did offer up patches to fix the issue, some really bad advice has been circulating the Internetz telling people to only support RC4, again. Sigh.
When RC4 finally dies a fiery death I’ll likely throw a party. I’m sure I won’t be the only one…
Due to a bug in mod_ssl, the ability to remove TLS 1.0 (and only support TLS 1.1 and/or TLS 1.2) has not been available. The fix has now made it to CentOS 6 and you can now fine-tune your cryptographic protocols with ease.
Before the fix my /etc/httpd/conf.d/ssl.conf file had this line:
SSLProtocol all -SSLv2 -SSLv3
This allows all SSL protocols except SSLv2 and SSLv3 to be used with httpd. This isn’t a bad solution but there are a couple of sites that I’d prefer to further lock down by removing TLS 1.0 and TLS
1.2 1.1. With the fix now in mod_ssl my settings can now look like this:
SSLProtocol all -SSLv2 -SSLv3 -TLSv1 -TLSv1.1
…and I’ll only support TLS 1.2 and beyond. Of course doing this will significantly reduce the number of clients that can connect to my server. According to SSLLabs I’m blocking all IE users before IE 11, Android before 4.4.2, Java 7, and Firefox 24.2.0 ESR. But luckily I really don’t have a problem with any of these browsers for a couple of things I do so I’ll likely tighten up security there and leave my more public sites alone.
NSS and mod_nss for httpd wasn’t discussed because it’s not in use on my systems. it should be noted that mod_nss can be similarly configured as mod_ssl however mod_nss does not support TLS 1.2 and you’ll max out at TLS 1.1.
My friend Hubert has been doing a lot of work to make better the world a little safer. Glad he’s getting some recognition. Here’s a great article on testing your server for proper SSL/TLS configurations.
I’ve been enjoying watching these trends.
Originally posted on securitypitfalls:
This time the results are not really different from past month’s ones. About two percent of servers more use SHA-256 signed certificates and 1% more has configuration that allows negotiation of PFS suites.
Small change to reported results: I’ve added “Insecure” entry which counts the number of servers that will use completely insecure cipher suite like single DES, RC2 or export grade ciphers. It doesn’t include the “controversial but not broken” IDEA and SEED ciphers.
SSL/TLS survey of 402742 websites from Alexa's top 1 million Stats only from connections that did provide valid certificates (or anonymous DH from servers that do also have valid certificate installed) Supported Ciphers Count Percent -------------------------+---------+------- 3DES 349454 86.7687 3DES Only 164 0.0407 AES 374868 93.0789 AES Only 1017 0.2525 AES-CBC Only 553 0.1373 AES-GCM 172322 42.7872 AES-GCM Only 7 0.0017 CAMELLIA 170577 42.3539 CHACHA20 15137 3.7585 Insecure 79666 19.7809 RC4 355750 88.332 RC4…
View original 1,216 more words