Bugzilla Bug 194236

iSEC Partners Security Advisory - 2007-006-RubySSL
http://www.isecpartners.com
--------------------------------------------

Ruby Net::HTTPS library does not validate server certificate CN

Vendor: Ruby
Vendor URL: http://www.ruby-lang.org
Versions affected: 1.8.5, 1.8.6, Trunk Ruby
Systems Affected: All Ruby Platforms
Severity: Medium - Compromise of SSL connection integrity
Author: Chris Clark <cclark[at]isecpartners[dot]com>

Vendor notified: Yes
Public release: Yes
Advisory URL: http://www.isecpartners.com/advisories/2007-06-rubyssl.txt

Summary:
--------
The Ruby Net::HTTP and Net::HTTPS libraries can be used to make HTTP or HTTPS 
connections to remote websites.  There are several methods for performing these 
types of connections within the Ruby standard library but these libraries
are recommended going forward.

A vulnerability results from the Net::HTTPS library failing to validate the 
name on the SSL certificate against the DNS name requested by the user.  By 
not validating the name, the library allows an attacker to present a 
cryptographically valid certificate with an invalid CN.  

Details:
--------
The vulnerability exists in the connect method within http.rb file which
fails to call post_connection_check after the SSL connection has been 
negotiated.  Since the server certificate's CN is not validated against 
the requested DNS name, the attacker can impersonate the target server 
in a SSL connection.  The integrity and confidentiality benefits of 
SSL are thereby eliminated.

Example:
If the application uses the following code to connect to the ip address of 
https://www.isecpartners.com:

url = "72.52.84.198" #www.isecpartners.com IP
path = "/"
http = Net::HTTP.new(url, 443)
http.use_ssl = true
http.ca_file = "verisign.pem"
http.verify_mode = OpenSSL::SSL::VERIFY_PEER
resp, data = http.get(path, nil)  

The connection will succeed.  This is an obvious failure as the certificate 
presented has a CN of www.isecpartners.com.  Mitigating factors are that
an attacker will have to be in a position to perform a middle person attack
and the attacker must possess a certificate signed by the CA specified in 
the ca_file attribute.  The latter is not a significant problem if 
targeting a site dependent on the public SSL infrastructure but may be
more difficult in cases where a private PKI is in use.

Fix Information:
----------------
This issue has been addressed by adding the appropriate post connection check 
within http.rb.  These patches add the enable_post_connection parameter, which 
if set to true, will cause the Net::HTTP library to raise an exception when the 
post connection check fails.

The following trunk Ruby checkin contains the related changes:
http://svn.ruby-lang.org/cgi-bin/viewvc.cgi?view=rev&revision=13499
http://svn.ruby-lang.org/cgi-bin/viewvc.cgi?view=rev&revision=13500

Ruby 1.8.5:
http://svn.ruby-lang.org/cgi-bin/viewvc.cgi?view=rev&revision=13502

Ruby 1.8.6:
http://svn.ruby-lang.org/cgi-bin/viewvc.cgi?view=rev&revision=13504

Thanks to:
----------
Rachel Engel, GOTOU Yuuzou, and Minero Aoki

About iSEC Partners:
--------------------
iSEC Partners is a full-service security consulting firm that provides
penetration testing, secure systems development, security education
and software design verification, with offices in San Francisco,
Seattle, Ewa Beach, and Los Angeles.

http://www.isecpartners.com
info@isecpartners.com

the correct url for the ipsecpartners advisory is
http://www.isecpartners.com/advisories/2007-006-rubyssl.txt

Richard, you bumped Ruby to 1.8.5_p113 and 1.8.6_p110-r1 a week ago. Is
1.8.5_p113 ok to go stable?
Can you remove non-stable vulnerable versions?

Rbu, sorry, I also added an apparently entirely imaginary comment to this bug
when I bumped them. 

x86 and amd64 are still on 1.8.5, they can either go to the latest 1.8.5 or
1.8.6, everyone else except mips is on 1.8.6. I haven't checked but I suspect
1.8.4 is also affected, so mips should probably stabilise a new vesrion as
well.

I've removed all the other versions

I was wrong, mips is on 1.8.6-r1 as well. Arches please stabilise

x86 stable

*** Bug 181110 has been marked as a duplicate of this bug. ***

ppc stable

alpha/ia64/sparc stable

Stable for HPPA.

amd64 stable for ruby-1.8.6-r1.ebuild

stable on ppc64:

dev-lang/ruby-1.8.5_p113
dev-lang/ruby-1.8.6_p110-r1

If someone calls arches he should mention the versions, which should go stable.

amd64, I think you marked the wrong version stable..

amd64 stable

time for glsa decision. I tend to vote NO.

I vote NO.

very rare - i vote no too. Closing with noglsa. Feel free to reopen if you
disagree.