Skip to navigation Skip to main content Skip to footer

Shellshock Advisory

25 September 2014

By Aaron Haymore

This research was originally performed by researchers from iSec Partners (now NCC Group), and has been migrated to research.nccgroup.com for posterity.

Shellshock Advisory

25 Sep 2014 – iSEC Partners

Executive Summary

Immediate patches are required to fix a vulnerability in bash that allows arbitrary code execution from unauthenticated users.

The full impact of vulnerable vectors may never be enumerated, so patching is required immediately, as in-the-wild attacks are being seen.

The vulnerability is not fully resolved by the available patch, so a second round of patching will be required once the subsequent patch is available.

It is not necessary to restart computers or processes following patching.

Impact Determining Exposure

Shellshock enables code execution, arbitrary file disclosure, and system compromise from unauthenticated remote attackers. The ways that a system could be vulnerable to this bug are numerous, and no exhaustive list could be compiled. However, some of the common scenarios are:

  • Web servers using CGI scripts that are written in bash
  • Web servers using CGI scripts that are written in other languages that invoke certain function calls:
    • C-based scripts calling system() or popen()
    • Python-based scripts that call os.system() or os.popen()
    • PHP-based scripts that call system() or exec() (when run in CGI mode)
    • Perl-based scripts that invoke shell commands
  • Restricted SSH shells using ForceCommand can be bypassed. Some git and subversion deployments use such restricted shells.
  • If an attacker is in a position to forge DHCP responses, it can enable root-level code execution in DHCP clients
  • Set-UID applications may allow local escalation to root
  • CUPS-based printer daemons are likely to be affected
  • Mac and Linux Desktops are affected, and are likely to allow privilege escalation to a root user

Certain common deployments are not affected:

  • Regular use of SSH is not affected as users already have shell access
  • PHP scripts that use mod_php are not affected, nor is mod_python or mod_perl
  • Sudo by itself is not affected

Finally, there is no need to restart services after patching. Running processes have passed the window of vulnerability and new processes will be running the new, patched code.

Patching

Linux Operating Systems have deployed patches:

Mac OS X is not yet patched, but manual instructions to recompile bash from source are available at: https://apple.stackexchange.com/questions/146849/

An Imperfect Fix

Unfortunately, the patch supplied is incomplete. As noted by Tavis Ormandy, other vectors still exist to bypass protections and perform invalid actions, such as overwriting files. While no one has publicly demonstrated code execution for this bypass, it seems likely to be possible. A second CVE (CVE-2014-7169) has been created to track this issue.

Red Hat has an experimental mitigation that requires many manual steps to use available at: https://access.redhat.com/articles/1200223

Red Hat and NCC both advise deploying the available patch immediately and being prepared to deploy a second patch, when one becomes stable and tested shortly.

In the Wild Attacks

Shellshock is being actively scanned for and exploited on the Internet at-large currently.

Robert Graham is one security researcher who has initiated an Internet-wide scan from the IP address 209.126.230.72. His scan, and many others, use a ping command to call back to a server, alerting them that your server is vulnerable – although this is not the only mechanism one could use.

Besides internet scans, several Metasploit modules are available, and a few exploits have been posted online. These include exploits that provide shell access to a server, read arbitrary files the web server has access to read, and a report of a payload that exploits a kernel vulnerability. The kernel exploit is not yet confirmed to exploit a previously known or unknown vulnerability.

IDS Signatures and Detection

IDS vendors are producing rules that will attempt to detect and block attempted exploitation of this issue. Rules are available for:

Technical Details of the Flaw

The vulnerability was discovered by Stephane Chazelas and can be tested for manually using the following command:

env x='() { :;}; echo vulnerable' bash -c "echo this is a test"

If it outputs ‘vulnerable’, the system is vulnerable. This bug arises because of an unusual feature of bash that allows exporting functions as well as environment variables. This feature is specific to bash, and no other shells are known to be vulnerable. (However, on some systems, other shells will actually be symlinks to bash.)

The actual vulnerability is in the parser for these exported functions. It does not parse the function correctly, and upon invocation will automatically execute trailing code defined after the function.

Any variable beginning with “() {“ is automatically treated as a function – but the aspect that makes this bug so prevalent is that environment variables are populated in unexpected places from user input. For example, environment variables like HTTP_COOKIE and HTTP_USER_AGENT are often populated for CGI scripts. And PHP, Perl, Python, and other scripts are often run as CGI scripts under a web server.

This results in a perfect storm of unexpected vectors of automatic remote code execution, most commonly on web servers. For more details, a good technical blog post is: http://lcamtuf.blogspot.co.uk/2014/09/quick-notes-about-bash-bug-its-impact.html

References