DNS Reverse Lookup as a vector for the Bash vulnerability (CVE-2014-6271 et.al.)
CVE-2014-3671
references:
CVE-2014-6271, CVE-2014-7169, CVE-2014-6277, CVE-2014-6278
CVE-2014-7186 and, CVE-2014-7187
* Summary:
Above CVEs detail a number of flaws in bash prior related to the parsing
of environment variables (aka BashBug, Shellshock). Several networked
vectors for triggering this bug have been discovered; such as through
dhcp options and CGI environment variables in webservers [1].
This document is to advise you of an additional vector; through a
reverse lookup in DNS; and where the results of this lookup are
passed, unsanitized, to an environment variable (e.g. as part of
a batch process).
This vector is subtly different from a normal attack vector, as the
attacker can 'sit back' and let a (legitimate) user trigger the
issue; hence keeping the footprint for a IDS or WAAS to act on small.
* Resolvers/systems affected:
At this point of time the stock resolvers (in combination with the libc
library) of OSX 10.9 (all versions) and 10.10/R2 are the only known
standard installations that pass the bash exploit string back and
up to getnameinfo().
That means that UNpatched systems are vulnerable through this vector
PRIOR to the bash update documented in http:
Most other OS-es (e.g. RHEL6, Centos, FreeBSD 7 and up, seem
unaffected in their stock install as libc/libresolver and DNS use
different escaping mechanisms (octal v.s. decimal).
We're currently following investing a number of async DNS resolvers
that are commonly used in DB cache/speed optimising products and
application level/embedded firewall systems.
Versions affected:
See above CVEs as your primary source.
* Resolution and Mitigation:
In addition to the mitigations listed in above CVEs - IDSes and similar
systems may be configured to parse DNS traffic in order to spot the
offending strings.
Also note that Apple DL1769 addresses the Bash issue; NOT the vector
through the resolver.
* Reproducing the flaw:
A simple zone file; such as:
$TTL 10;
$ORIGIN in-addr.arpa.
@ IN SOA ns.boem.wleiden.net dirkx.webweaving.org (
666 ; serial
360 180 3600 1800 ; very short lifespan.
)
IN NS 127.0.0.1
* PTR "() { :;}; echo CVE-2014-6271, CVE-201407169, RDNS"
can be used to create an environment in which to test the issue with existing code
or with the following trivial example:
#include <sys/socket.h>
#include <netdb.h>
#include <assert.h>
#include <arpa/inet.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <netinet/in.h>
int main(int argc, char ** argv) {
struct in_addr addr;
struct sockaddr_in sa;
char host[1024];
assert(argc==2);
assert(inet_aton(argv[1],&addr) == 1);
sa.sin_family = AF_INET;
sa.sin_addr = addr;
assert(0==getnameinfo((struct sockaddr *)&sa, sizeof sa,
host, sizeof host, NULL, 0, NI_NAMEREQD));
printf("Lookup result: %s\n\n", host);
assert(setenv("REMOTE_HOST",host,1) == 0);
execl("/bin/bash",NULL);
}
Credits and timeline
The flaw was found and reported by Stephane Chazelas (see CVE-2014-6271
for details). Dirk-Willem van Gulik (dirkx(at)webweaving.org) found
the DNS reverse lookup vector.
09-04-2011 first reported.
2011, 2014 issue verified on various embedded/firewall/waas
systems and reported to vendors.
??-09-2014 Apple specific exploited seen.
11-10-2014 Apple confirms that with DL1769 in place that
"The issue that remains, while it raises
interesting questions, is not a security
issue in and of itself."
* Common Vulnerability Scoring (Version 2) and vector:
See CVE-2014-6271.
1:https:
1.10 / : 1726 $
