Re: Dealing with compromised Zimbra accounts

[Matt Mencel <MR-Mencel@wiu.edu>]

My cohort can answer better than me as he's done most of the work concerning this.....but...we've kind of discovered a few things about spammers using our Zimbra...

1)  The spammers don't waste much time sending single messages, except while testing which maybe helps our little trap.
2)  The spammers automation stuff only seems to work in the HTML(Standard) client.

So, when we find an outgoing message with an envelope that contains more than X (I think 50) recipients AND it's been sent from the HTML(Standard) client, we dump it into a queue and don't deliver until it's manually reviewed by an administrator.

I'd guess we're catching almost 100% of the outgoing SPAM attempts now when an account gets compromised.  The AJAX client may be too difficult for them to script around(I'm guessing), and they like to send to LOTS of recipients.  We've only found one or two valid senders who have been caught so far....and they've either switched back to the AJAX client or understand why we delay their multi-recipient messages.

Phishing is one of those issues that separates the regular people from the idiots....and we seem to have our fair share of the latter here.  Not only do they continue to respond after lots of training and high profile example cases....but they respond when the message is delivered TO THE JUNK!! folder...AND...the subject line INCLUDES some text that reads [Fraud Alert: Suspected SPAM]....

HELLOOO???  The lights are on but no one is home.....I imagine the thought process goes like this...

....must...fight...urge....but...I...have....no....common...sense...
must...hit..."Reply" button....and...give...away...my...private...info...


It's cause for much consternation....but we do get a few laughs out of it sometimes.  Maybe immense public ridicule would work?  :)

Matt



----- Original Message -----
From: "Tom Golson" <tgolson@tamu.edu>
To: "Steve Hillman" <hillman@sfu.ca>
Cc: "zimbra-hied-admins" <zimbra-hied-admins@sfu.ca>
Sent: Tuesday, January 13, 2009 2:41:10 PM GMT -06:00 US/Canada Central
Subject: Re: Dealing with compromised Zimbra accounts

It has been our experience that setting the zimbraAccountStatus to
"locked" does kill the current session.  But you're right -- if used as
a "toggle" there's no useful effect.  The mailbox has to remain locked
long enough for the spammer to lose interest.

We use a semi-automated process of monitoring outbound message rates and
triggering alerts.  Since our customers do have the unfortunate habit of
 sending mail to massive lists of people, we try to verify messages as
being "spam" before we take punitive action.

If an account is compromised, one of our admins will trigger a process
to scramble the customer's password, lock their mailbox and send a
notice to our help desk so they can conduct "customer education".  We do
leave the mailbox locked until the customer has been contacted and reset
their password.

Inbound, we make very heavy use of the anti-phishing-mail-reply group
that was spun off from the hied-email-admin list onto Google groups.  It
is an excellent resource.

--Tom

Steve Hillman wrote:
> Hi folks,
>   Now that we've moved all of our students over to our Zimbra implementation, we're starting to see Zimbra accounts get used for spamming (after being successfully phished) -- we've had 3 in the last 3 days.
> 
> We're having to invent new ways of dealing with these -- simply changing the password or locking the account doesn't actually stop the current session. Our session timeout is 12 hours, so this would allow the spammer to keep functioning for quite awhile unless we can kill the session programmatically. 
> 
> I've found that setting the account to 'maintenance' mode will terminate the session *if* the end-user tries to do anything while it's in that state, but it's not enough to just flip the account into maintenance and back out.
> 
> I'm just wondering if any other sites are dealing with this yet, and if so, if you've figured out a semi-automated way of locking and unlocking the accounts?
> 
> (btw, we're also looking at Milters to do password detection (to prevent the phished account in the first place) and outbound rate limiting (to limit the damage the spammer can do), so if any of you are doing/done any work in this area, I'd love to hear about it too..)
>