Educause Security Discussion mailing list archives

Re: Spammer got into my Webmail


From: Robin Polak <robin.polak () GMAIL COM>
Date: Tue, 9 Sep 2008 11:33:37 -0400

That is really helpful.  We are also using Mimedefang.  How is that you got
it to make that comparison and log the results?

On Tue, Sep 9, 2008 at 11:26, Joel Rosenblatt <joel () columbia edu> wrote:

As to the IMP question:

To insert the sender in IMP, see imp/config/header.php.  Our code
is one line:

$_header['X-CubMail-Sender'] = $GLOBALS['imp']['uniquser'];

CubMail is the name of our IMP installation, so change that :-)


We use a very flexible sendmail milter called Mimedefang, and we
instruct it to compare the user in the X-CubMail-Sender line to the
address in the From line, and log them and the number of recipients
if they don't match.  No alarm is sent since messages like this are
usually OK.  But if we have an abuse report, even with incomplete
headers, we can figure out what account was used.

Thanks,
Joel Rosenblatt

Joel Rosenblatt, Manager Network & Computer Security
Columbia Information Security Office (CISO)
Columbia University, 612 W 115th Street, NY, NY 10025 / 212 854 3033
http://www.columbia.edu/~joel <http://www.columbia.edu/%7Ejoel>



--On Tuesday, September 09, 2008 11:01 AM -0400 Robin Polak <
robin.polak () GMAIL COM> wrote:

 What did you put into your header.txt to set the X-Originating-User
header?
Would you be at all willing to share those scripts you use to monitor your
e-mail?  Thank You for your help!!

On Tue, Sep 9, 2008 at 10:49, Mark Montague <markmont () umich edu> wrote:

 Our Horde/IMP installation sets the X-Originating-User header in outgoing
messages that we can use to identify messages in our outbound queue from
the
compromised account.

We have a script that runs periodically that monitors the size of our
outgoing webmail queues and notifies us when one gets unexpectedly large.
 Our experience has been that many spammers using compromised accounts
will
attempt to send thousands of messages in a very short period of time,
which
triggers a notification from our script and we can log in and see if
there
is actually a compromised account.

Our experience has also been that spammers using compromised accounts
with
Horde/IMP change the user's signature to contain the spam that they want
to
send.   Horde/IMP automatically fills in each message with the spam via
the
signature.  These spammers set up an additional identity for the user
with a
"From" address that is actually the spammer's address.  So we have a
script
that looks for new identities that get set up with non-local identities
and
emails us the first 100 characters of the signature for that identity --
this is enough to let us identify accounts that have been compromised,
sometimes soon enough that we can disable the account before the spammer
actually starts sending their messages.

Finally, we have a php_include file that we set up for Horde/IMP that
contains a blacklist of compromised user accounts.  We can use this to
block
a compromised account from using webmail without having to completely
disable the account across all services.

             Mark Montague
             ITCS Web/Database Production Team
             The University of Michigan
             markmont () umich edu





On Tue, Sep 9, 2008 10:16 AM, Dan Oachs <doachs () GAC EDU> wrote:

 We have configured our Horde/IMP installation to use smtp
authentication.
 Our postfix logs then show who authenticated each message.  We can then
use
that information to remove messages from our outbound queue.

--Dan Oachs
 Gustavus Adolphus College



Robin Polak wrote:

 Hello,

 One of my webmail users was fooled into revealing his credentials to a
spammer and now I am dealing with the backlash of all this spam having
left
our smtp servers as well as much mail still left in the outbound
sendmail
queues.  Is there any advice that any of you could provide me as far as
filtering out the spam from my sendmail queues as well as any
procedures I
could follow to counteract the effects of blacklisting such as a
generally
checked whitelist?  In addition, as a result of this incident I have
found a
flaw in the tracking of mail between our webmail (Horde/IMP), Cyrus
IMAP,
and Sendmail.  What sort of suggestion could be made as far as
effectively
being able to correlate my logs?  Is there a way to put a header into a
message leaving IMP indicating the user-name that was used to login to
Horde?  This would have been quite usefull since in some way the
spammer was
able to spoof the From address in the message to be a yahoo.com <
http://yahoo.com>  address.

--
Robin Polak, Network Manager
College of Mount Saint Vincent
E-Mail: robin.polak () gmail com <mailto:robin.polak () gmail com>
V. 718-405-3293




--
Robin Polak
E-Mail: robin.polak () gmail com
V. 917-494-2080




Joel Rosenblatt, Manager Network & Computer Security
Columbia Information Security Office (CISO)
Columbia University, 612 W 115th Street, NY, NY 10025 / 212 854 3033
http://www.columbia.edu/~joel <http://www.columbia.edu/%7Ejoel>




--
Robin Polak
E-Mail: robin.polak () gmail com
V. 917-494-2080

Current thread: