MITKRB5-SA-2008-001: double-free, uninitialized data vulnerabilities in krb5kdc

Tue Mar 18 14:28:42 EDT 2008

CVE-2008-0062
VU#895609
Use of a null or dangling pointer in the MIT Kerberos KDC can result
in a crash or double-free, and may leak portions of process memory to
an attacker.

CVSSv2 Vector: AV:N/AC:M/Au:N/C:P/I:P/A:C/E:P/RL:O/RC:C

CVSSv2 Base Score:      9.3

Access Vector:          Network
Access Complexity:      Medium
Authentication:         None
Confidentiality Impact: Complete
Integrity Impact:       Complete
Availability Impact:    Complete

CVSSv2 Temporal Score:  6.5

Exploitability:         Proof-of-Concept
Remediation Level:      Official Fix
Report Confidence:      Confirmed

CVE-2008-0063
VU#895609

Uninitialized stack values cause re-use of a small window of previous
stack values to be interpreted as message content.  Some of the
"content" may be returned to the attacker as part of an error
response.

CVSSv2 Vector: AV:N/AC:M/Au:N/C:P/I:N/A:N/E:P/RL:O/RC:C

CVSSv2 Base Score:      4.3

Access Vector:          Network
Access Complexity:      Medium
Authentication:         None
Confidentiality Impact: Partial
Integrity Impact:       None
Availability Impact:    None

CVSSv2 Temporal Score:  3.4

Exploitability:         Proof-of-Concept
Remediation Level:      Official Fix
Report Confidence:      Confirmed

SUMMARY
=======

When Kerberos 4 support is enabled in the MIT Kerberos 5 KDC,
malformed messages may trigger two bugs:

CVE-2008-0062: A global variable holding a pointer to the message to
be sent back to the client is only set for two recognized krb4 message
types, but may be used (and freed) in additional cases, resulting in
use of a null or dangling pointer.

CVE-2008-0063: The incoming krb4 message is copied into a fixed-size
buffer on the stack, but the remainder of the buffer is left
untouched, and the bounds checks use the size of the buffer, not the
size of the data copied into it.

By default, Kerberos 4 support is compiled in but not enabled in
recent versions, and these bugs are not exposed unless Kerberos 4
support is enabled.

These are implementation bugs, not protocol defects.

IMPACT
======

CVE-2008-0062: An unauthenticated remote attacker may cause a
krb4-enabled KDC to crash, expose information, or execute arbitrary
code.  Successful exploitation of this vulnerability could compromise
the Kerberos key database and host security on the KDC host.

CVE-2008-0063: An unauthenticated remote attacker may cause a
krb4-enabled KDC to expose information.  It is theoretically possible
for the exposed information to include secret key data on some
platforms.

AFFECTED SOFTWARE
=================

MIT Kerberos 5 version 1.6.3 KDC, and probably all earlier versions,
when krb4 support is compiled in and enabled.  (The krb4 support is
disabled by default in recent releases.)  No client or application
server programs are affected.

FIXES
=====

* Apply the patch available at:
  http://web.mit.edu/kerberos/advisories/2008-001-patch.txt
  or in PGP-signed form at:
  http://web.mit.edu/kerberos/advisories/2008-001-patch.txt.asc

* These bugs will be fixed in the next release.

REFERENCES
==========

This announcement is posted at:

  http://web.mit.edu/kerberos/advisories/MITKRB5-SA-2008-001.txt

This announcement and related security advisories may be found on the
MIT Kerberos security advisory page at:

        http://web.mit.edu/kerberos/advisories/index.html

The main MIT Kerberos web page is at:

        http://web.mit.edu/kerberos/index.html

CVSSv2:

    http://www.first.org/cvss/cvss-guide.html
    http://nvd.nist.gov/cvss.cfm?calculator&adv&version=2

CVE: CVE-2008-0062 CVE-2008-0063
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2008-0062
http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2008-0063

CERT: VU#895609
http://www.kb.cert.org/vuls/id/895609

CONTACT
=======

The MIT Kerberos Team security contact address is
<krbcore-security at mit.edu>.  When sending sensitive information,
please PGP-encrypt it using the following key:

pub   1024D/2915318C 2008-01-18 [expires: 2009-02-01]
uid     MIT Kerberos Team Security Contact <krbcore-security at mit.edu>
sub   2048g/3A91A276 2008-01-18 [expires: 2009-02-01]

DETAILS
=======

CVE-2008-0062: If a bogus Kerberos 4 message (i.e., a message with the
first byte having the value 4, but the second byte not describing one
of the message types supported by the KDC) is received by the KDC, and
there has been no previous Kerberos 4 traffic, a null pointer
dereference will result, likely crashing the KDC.  If there has been
valid Kerberos 4 traffic already, a dangling pointer will be used to
locate the message to send to the client; it may resend a previously
generated response, send some other arbitrary chunk of process memory,
perhaps including secret key data, or crash the process by attempting
to access an invalid address.  If the process doesn't crash, random
addresses will be passed to free(), likely corrupting the free pool,
and potentially leading to later crashes, data corruption, jumps to
arbitrary locations in process memory, etc.

The KDC normally runs without write access to its database, so it is
not likely to corrupt the database, except insofar as arbitrary code
execution could theoretically corrupt anything the process has access
to on the system.

CVE-2008-0063: If a Kerberos 4 message is truncated, previous contents
of the stack may be used in place of the missing portions of the
message.  (Note that if the message type is missing, and the data read
from the stack is not a recognized message type, this may indirectly
trigger CVE-2008-0062 described above.)  Several strings are read from
the "message" as parts of principal names; these strings are limited
to 40 bytes or the next ASCII NUL found in the buffer.  If the KDC
returns an error message indicating that a principal name is not found
in its database, the principal name is included in the error message,
and some of the old stack content may be there.

If the previously handled message was a valid Kerberos 4 message,
parts of that message may be re-used for the new message; this
wouldn't expose any data that wouldn't have been visible on the local
network.

If the previously handled message was a Kerberos 5 message, the values
overlaid by the buffer are likely to be old argument pointers, saved
registers, return addresses, and so forth.  However, since stack
contents and layout are highly dependent on the platform and compiler,
it is impossible to assert that no secret key data may be leaked into
the exposed stack regions on any platform.

REVISION HISTORY
================

2008-03-18      original release

Copyright (C) 2008 Massachusetts Institute of Technology

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iD8DBQFH4AC6UqOaDMQ+e5gRAt5BAKCkfIKFE6assZ+fhbf8ghT5PsS5RQCfcQAJ
MmnThImfNYzxigqYCX+Fkm8=
=zmzD
-----END PGP SIGNATURE-----