Onderwerp: IMP 2.x SQL injection vulnerabilities

IMP is a popular webmail package written in PHP. It ships with some UNIX
systems and is also used on Windows servers. The version 2 of the program
contains some SQL injection flaws which allow any remote user to access
the webmail system's database. Valid user authentication is not required
in order to exploit the flaws.

The error happens in some database functions in PHP files named
lib/db.<databasename>. An example from db.pgsql, function check_prefs:

$sql="select username from $default->db_pref_table where username='$user@$server'";

Including user-supplied strings directly in an SQL query is a mistake.
The fix is to use something like the addslashes() PHP function.

As a proof of concept:

$ lynx "http://webmail.server/imp/mailbox.php3?actionID=6&server=x&imapuser=x';somes ql+--&pass=x"

IMP would try to execute "somesql" and the result would be this kind
of PHP error (presuming the PHP configuration allows displaying error
messages on web pages):

Warning: PostgreSQL query failed: ERROR: parser: parse error at or near "somesql" in
/usr/share/horde/imp/lib/db.pgsql on line 127

Even though SQL query results aren't directly readable from the screen
in the above example, the attacker might e.g. update his/her mail
signature to contain wanted query results and then view it on the
preferences page of IMP. This requires a valid login, but isn't a
problem for an attacker because IMP allows the use of any remote IMAP
server. Use of the server_list option doesn't affect this behaviour; the
attacker-controlled IMAP server may be still passed to mailbox.php3 in the
URL.

The impact of SQL injection depends heavily on the underlying database
and its configuration. If PostgreSQL is used, it's possible to execute
multiple complete SQL queries separated by semicolons. The database
contains session id's so the attacker might hijack sessions of people
currently logged in and read their mail. In the worst case, if the
hordemgr user has the required privilege to use the COPY SQL command
(found in PostgreSQL at least), a remote user may read or write to any
file the database user (postgres) can. The attacker may then be able to
run arbitrary shell commands by writing them to the postgres user's
~/.psqlrc; they'd be run when the user starts the psql command which
under some configurations happens regularly from a cron script.

If other database servers are used, the exploitation possibilities may
be more limited.

The vendor has been informed about this bug last month. Although there
hasn't been any direct reply, there was a comment on this on the IMP
mailing list: "2.2.x is officially deprecated/unsupported. This does not
apply to 3.x.".

Versions up to and including 2.2.8 seem vulnerable. According to the
author, version 3 isn't affected so upgrading to IMP 3 is recommended.
This, and more information about IMP is available at http://horde.org/imp/.



Jouko Pynnönen
jouko@solutions.fi

On Wed, 8 Jan 2003, Jouko Pynnonen informed us that:

> The vendor has been informed about this bug last month. Although there
> hasn't been any direct reply, there was a comment on this on the IMP
> mailing list: "2.2.x is officially deprecated/unsupported. This does not
> apply to 3.x.".
>
> Versions up to and including 2.2.8 seem vulnerable. According to the
> author, version 3 isn't affected so upgrading to IMP 3 is recommended.
> This, and more information about IMP is available at http://horde.org/imp/.

What many software developpers (including, but apparently not limitted
to, many commercial software vendors) seem to fail to realize is that
some sites use their applications in production environments, with (in
my case tens of thousands of) real users and upgrading to the latest
version which includes numerous changes above and beyond the fix for the
reported bug is often difficult in the best of cases.

In the case of Imp-2 -> Imp-3, the changes are much too significant for
some of us to simply switch versions and hope our user community doesn't
notice. It's a lot easier for us to patch-in-place to deal with the bug
itself, and leave the upgrade to new features (and new bugs!) to be done
in a more coordinated fashion, with time for users to evaluate the new
interface, etc.

That being said, and thanks to the information in Jouko's advisory,
I've patched our own Imp installation (which has now had so many patches
applied I should start giving it local version numbers!) according to
the appended. I hope others will be able to make use of this patch,
and I especially hope that if I've overlooked something, others will
point it out...

--
----------------------------------------------------------------------
Sylvain Robitaille syl@alcor.concordia.ca

Systems analyst Concordia University
Instructional & Information Technology Montreal, Quebec, Canada
----------------------------------------------------------------------

# Of course, folks using Imp-2 with non-PostgreSQL databases will
# need to adapt the following to the appropriate db.* file

--- lib/db.pgsql.20030108 2000-12-20 15:45:33.000000000 -0500
+++ lib/db.pgsql 2003-01-08 15:18:25.000000000 -0500
@@ -26,6 +26,13 @@
function imp_add_address ($address, $nickname, $fullname, $user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $address = addslashes($address);
+ $nickname = addslashes($nickname);
+ $fullname = addslashes($fullname);
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: adds $address, $nickname, $fullname to the addressbook for $user@$server
returns true on success and false on failure
*/
@@ -41,6 +48,10 @@
function imp_check_prefs ($user, $server) {
global $_imp_prefs_exist, $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $user = addslashes($user);
+ $server = addslashes($server);
+
if (isset($_imp_prefs_exist)) {
return $_imp_prefs_exist;
}
@@ -59,6 +70,11 @@
function imp_delete_address ($address, $user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $address = addslashes($address);
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: deletes $address from the addressbook of $user@$server
returns true on success and false on failure
*/
@@ -72,6 +88,10 @@
function imp_get_addresses ($user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: returns a 2d array of addresses where each
element is an array in which element 0 is the address,
element 1 is the nickname, and element 2 is the fullname.
@@ -92,6 +112,10 @@
function imp_get_from ($user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: returns the signature for the database key $user@$server
(a string), or false on failure.
*/
@@ -105,6 +129,10 @@
function imp_get_fullname ($user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: returns the signature for the database key $user@$server
(a string), or false on failure.
*/
@@ -118,6 +146,10 @@
function imp_get_lang ($user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: returns the signature for the database key $user@$server
(a string), or false on failure.
*/
@@ -131,6 +163,10 @@
function imp_get_signature ($user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: returns the signature for the database key $user@$server
(a string), or false on failure.
*/
@@ -144,6 +180,11 @@
function imp_set_from ($from, $user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $from = addslashes($from);
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: sets the replyto to $from for the database key $user@$server
returns true on success and false on failure
*/
@@ -165,6 +206,11 @@
function imp_set_fullname ($fullname, $user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $fullname = addslashes($fullname);
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: sets the fullname to $fullname for the database key $user@$server
returns true on success and false on failure
*/
@@ -186,6 +232,11 @@
function imp_set_lang ($lang, $user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $lang = addslashes($lang);
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: sets the language to $lang for the database key $user@$server
returns true on success and false on failure
*/
@@ -208,6 +259,11 @@
function imp_set_signature ($signature, $user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $signature = addslashes($signature);
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: sets the signature to $signature for the database key $user@$server
returns true on success and false on failure
*/
@@ -230,6 +286,14 @@
function imp_update_address ($old_address, $address, $nickname, $fullname, $user, $server) {
global $default;

+ /* 2003/01/08 Sylvain Robitaille: Sanitize our input. */
+ $old_address = addslashes($old_address);
+ $address = addslashes($address);
+ $nickname = addslashes($nickname);
+ $fullname = addslashes($fullname);
+ $user = addslashes($user);
+ $server = addslashes($server);
+
/* post: changes the entry for $old_address to $address, $nickname, $fullname.
returns true on success and false on failure
*/