Issue #5418 has been updated by naruse (Yui NARUSE).

Assignee changed from nahi (Hiroshi Nakamura) to normalperson (Eric Wong)

As Rails did, webrick seems to need introduce TRUSTED_PROXIES.

----------------------------------------
Bug #5418: Some properties of WEBrick::HTTPRequest could be malformed
https://bugs.ruby-lang.org/issues/5418#change-73426

* Author: nahi (Hiroshi Nakamura)
* Status: Assigned
* Priority: Normal
* Assignee: normalperson (Eric Wong)
* Target version: 
* ruby -v: -
* Backport: 
----------------------------------------
Original reported issue: CVE-2011-3187

Users may expect that properties of WEBrick::HTTPRequest to be not
malformed/faked. But at the fact, in current implementation, following
properties can be malformed and faked by HTTP header sent by attacker.

 - HTTPRequest#host
  - can be malformed/faked by 'x-forwarded-host'
  - can be faked by 'Host'

 - HTTPRequest#port
  - can be faked by 'Host'

 - HTTPRequest#server_name
  - can be malformed/faked by 'x-forwarded-server'

 - HTTPRequest#remote_ip
  - can be malformed/faked by 'x-forwarded-for' and 'client-ip'

 - HTTPRequest#ssl?
  - can be faked by 'Host'

 - HTTPRequest#meta_vars (Hash of meta vars such as 'REQUEST_URI')
  - can be malformed/faked by some HTTP headers

Here's the list of reason why we're thinking it's not a
high-priority security bug at this moment.

 - For faked data issue, we don't have a way to guarantee that it's not
  faked. So developers of HTTPRequest must aware of that.

 - For malformed data issue, it should be a bug of HTTPRequest to be
  fixed, but it's the same problem for x-forwarded-host,
  x-forwarded-server and client-ip. We're offering those data in as-is
  basis from HTTP header so we can expect users handle the data
  properly for their purpose (for dumping to xterm, embedding to HTML,
  etc.)

 - And the fix for this bug would be a little complex for quick-fix
  because it's not only x-forwarded-for which causes this issue.
  'client-ip' needs care, too. Documentation would be enough for
  server_name. We think we need general development cycle for fixing
  it.

ref:
https://bugzilla.novell.com/show_bug.cgi?id=673010
http://webservsec.blogspot.com/2011/02/ruby-on-rails-vulnerability.html




-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>