ISC Diary

Refresh Latest Diaries

Handler on Duty:
Kevin Shortt
Contact Us

SSL/TLS Vulnerability Details to be Released Friday

Published: 2011-09-20,
Last Updated: 2011-09-20 15:18:13 UTC
by Kevin Liston (Version: 2)

I'm getting a lot of emails asking about articles that ultimately reference this upcoming talk: "BEAST: Surprising crypto attack against HTTPS" (http://ekoparty.org/2011/juliano-rizzo.php)

I don't have any extra details. Anything that I write now will be unnecessary speculation. It sounds like it will be interesting; their presentation last year on Padded Oracle Attacks (the crypto Oracle, not the database) certainly was.

UPDATE: Dr J links us to "A CHALLENGING BUT FEASIBLE BLOCKWISE-ADAPTIVE
CHOSEN-PLAINTEXT ATTACK ON SSL" that may describe the attack. This attack requires that the attacker be able to sniff the traffic and run code on the victims machine to inject the chosen-plaintext into the stream.

My recommendation is still to wait until we see the details before formulating a response, but sight-unseen the following steps couldn't hurt:

Users: Don't bank using someone else's wifi.
Browser Authors: Update to support TLS 1.2
Servers Admins: Configure to support TLS 1.2

Keywords: TLS

Comments

Juliano Rizzo presents this friday the attack and the tool "Beast" at Ekoparty.

More info here:
https://threatpost.com/en_us/blogs/new-attack-breaks-confidentiality-model-ssl-allows-theft-encrypted-cookies-091611

http://www.theregister.co.uk/2011/09/19/beast_exploits_paypal_ssl/

posted by perox_, Tue Sep 20 2011, 07:27

Yes, those would be the articles. I'd like to postpone any response until we see the actual presentation.

-KL

posted by Kevin, Tue Sep 20 2011, 13:00

an older paper that appears to describe this type of attack. I am with Kevin on withholding further discussion until seeing the attack/paper: http://eprint.iacr.org/2006/136.pdf

posted by jullrich@sans.edu, Tue Sep 20 2011, 13:27

So, I just know my boss is going to ask,. "If SSL is dead and TLS is broken, how do we secure our web sites?" I don't have a good answer for him. Thoughts?

posted by RobM, Tue Sep 20 2011, 14:19

@RobM's boss: we recommend waiting until Friday when there are details to analyze. But if you insist on doing something now, upgrading your servers to support TLS 1.2 couldn't hurt.

posted by Kevin, Tue Sep 20 2011, 15:09

Anyone have a link on how to upgrade Apache 2.2 to support TLS 1.2? The apache 2.2 reference here: http://httpd.apache.org/docs/2.2/mod/mod_ssl.html#sslprotocol doesn't seem to provide a 1.2 protocol configuration option.

posted by Bob Stangarone, Tue Sep 20 2011, 19:32

I once attended a demonstration of Zcaler (hosted web filtering service) where they bragged about "breaking SSL on the fly" to monitor for data leak. Am I missing something here? Should the padlock icon be replaced by someone laughing?

posted by Confused, Wed Sep 21 2011, 00:48

Last comment should say "ZScaler" of course

posted by Confused, Wed Sep 21 2011, 00:48

@confused. They are using HTTP inspect, where you are presented a certificate from the device intercepting your request, with some information from the original cert.

Essentially they do a MITM on your SSL traffic. So not breaking the protocol at all.
M

posted by Mark, Wed Sep 21 2011, 05:08

openssl doesn't support TLS 1.1 or 1.2. Apache doesn't support TLS 1.1 or 1.2.. the browsers don't TLs 1.1 or 1.2.

This problem has been kept really quiet ... I wonder why

https://bugzilla.mozilla.org/show_bug.cgi?id=565047 � Implement TLS 1.1 (RFC 4346)
https://bugzilla.mozilla.org/show_bug.cgi?id=480514 � Implement support for TLS 1.2 (RFC 5246)

posted by Paradiso, Wed Sep 21 2011, 05:12

@Paradiso

Apache (on Unix) is able to do both TLS 1.1 and TLS 1.2 if you are using mod_gnutls instead of mod_ssl (so use GnuTLS instead of OpenSSL). I did some tests a while ago and it worked fine with IE9 and Opera.

On Windows, it may be able to do so out of the box if httpd is linked against the Microsoft crypto APIs. I've seen some entries on SSLLabs, but the server may have been using mod_gnutls too.

As for OpenSSL, the still-to-be-released "stable" 1.0.1 branch (see CVS) does at least TLS 1.1, tested against Nginx 1.0.x. I don't know how complete the implementation is, but it does successfully pass the SSLLabs scan.

Also, see the following thread in the OpenSSL-dev mailing list:

http://www.mail-archive.com/openssl-dev@openssl.org/msg29714.html

posted by Steve, Wed Sep 21 2011, 11:12

@Steve
I believe windows crypto is tied to the OS version. From comments I've been reading about this they're only present by default on Windows 7+ and Windows Server 2k8. I haven't tried but it's apparently possible to manually install them on Vista but not on XP.

posted by MCartegan, Wed Sep 21 2011, 23:11

Well since everyone else is speculating ... This attach seems to be against SSL using a block cipher, such as AES. While that's not good, lots of the web servers I've seen choose RC4 (which is a stream cipher) as their first choice when protecting their SSL traffic. So while disabling AES isn't a great answer (and using it might be a requirement for government sites), RC4 is a great algorithm (secure, fast), very widely adopted, and may not be vulnerable to this attack.

posted by Dave O, Thu Sep 22 2011, 14:47

New Comments closed for all Diaries older than two(2) weeks
Please send your comments to our Contact Form

Diary Archives