Bugtraq mailing list archives

Android DNS poisoning: Randomness gone bad (CVE-2012-2808)


From: Roee Hay <roeeh () il ibm com>
Date: Tue, 24 Jul 2012 15:35:17 +0300

1 Introduction
===========
Recently we discovered a very interesting vulnerability in Android’s
DNS resolver,
a weakness in its pseudo-random number generator (PRNG), which makes
DNS poisoning
attacks feasible.

The full advisory can be found at http://bit.ly/MkteBx
A blog post can be found at http://bit.ly/MkoU5j
Demo of our PoC can be found at http://youtu.be/ffnF7Jej7l0

2 Vulnerability
============
The PRNG that the DNS resolver uses is
        random_id = 0xffff & (time_usec ^ time_sec ^ pid)

where time_sec is the current time in seconds, time_usec is the microseconds
fraction and pid is the process identifier.

Both the TXID and source port are generated by this PRNG.
Since both calls occur subsequently, the values are very much correlated
to each other. This yields a feasible attack expected time as we show
that the number of
random bits is brought down from 32 (ideally) to less than 21.

Check our advisory for full details.

3 Vulnerable versions
================
Android 4.0.4 and below.

4 Vendor Response
===============
Android 4.1.1 has been released, and patches are available on AOSP. The random
sample is now pulled from /dev/urandom, which should have adequate entropy by
the time network activity occurs.

5 Identifier
========
CVE-2012-2808

6 Discovered by
============
Roee Hay & Roi Saltzman
IBM Application Security Research Group

7 Disclosure timeline
================
07/24/2012 Public disclosure
06/05/2012 Issue confirmed by Android Security Team and patch provided
to partners.
05/21/2012 Disclosed to Android Security Team.


Current thread: