A NIMDA Experience
by Ryan Smith
On Tuesday, September 18th 2001 a new worm was found that spread quickly over the entire world. Nimda
is a complex virus with a mass mailing worm component which spreads itself very rapidly. In an effort
to give insight into some of what a network administrator does, here is a recount of my experiences
with the NIMDA virus.
I do know that after 9:34 am, this virus began propogating itself to directories on the local
server in the form of .EML files.
9:53 am
10:13 am
10:17 am
10:18 am
10:22 am
At this point, I connected to the D$ share of the IIS/Home server and began a File Find for *.EML. I
also started a full virus scan of the server using InoculateIT (running the latest signature's of course)
10:42 am
10:54 am
11:27 am
11:58 am
12:06 pm
12:21 pm
12:24 pm
12:29 pm
12:41 pm
12:52 pm
We spend the next hour and a half deleting over 35,000 .EML files on five different servers.
2:30 pm
We're still combating the re-appearance of thousands of .EML files on several servers. As fast
as we can delete the files, something is creating more. We're convinced that we have users
desktop PC's that are infected (at least one, that's all it needs)
3:15 pm
We're continuing to delete thousands of .EML files on the servers, which thankfully aren't
re-appearing anymore now, while the desktop support group begins the arduous process of
manually checking over a hundred desktop PC's for any .EML files.
5:00 pm
8:00 pm
11:18 pm
While the signatures we're updating, I did a quick File Find on the IIS/Home server to see if
any .EML files popped up overnight. NOOO! There's over 5,000 .EML files back on this server!
We get approval from management to keep users at our location off of their computers again until
we can get this contained. It's not fun, but the majority of them understand. At least users
at our other three locations can still use their computers, but the Internet and Internet e-mail
is still down. We spend the next 5 hours cleaning servers and continuing to clean user's desktop
PC's. We also begin applying IE 5.01 SP2 patches to desktops
12:25 pm
We decide that this server is beyond recovery. It's going to need a complete system rebuild and
data needs to be recovered back from Monday's back-up (pre-NIMDA) tape. We isolate the computer
from the network and spend the next three hours documenting every single piece of information we
can from the server, to make the rebuild easier.
1:45 pm
4:15 pm
7:45 am
8:30 am
12:15 pm
2:10 pm
Roughly two and a half-days later and a completely rebuilt server, we're back online. So what did we
learn from this? Well, our systems are certainly kept up to date with hotfixes and security patches
now. We always tried to stay a few months behind the curve, to let the other guys test things out first.
No longer the case, we're not going to be the early-adopters. To the extent that we're gambling the
stability of our systems in an effort to prevent the next big NIMDA virus from hitting us. However,
for us that's a gamble that we've decided we're willing to take to ensure that we don't spend another
two and a half-days dealing with this problem next time. But then again, no matter how many times
you think that's going to happen, it never seems to work does it.
September 18th
9:34 am
Unbeknownst to us, our internal IIS server (which is also a HOME server for several users) used
for running Microsoft OWA (Exchange's Outlook Web Access) is affected by a Unicode exploit from an
outside source and is infected by the NIMDA virus. At this time, no one knew what this virus was or
what it did. It didn't even have a name for a few more hours.
The NIMDA (admin backwords) worm spread quickly over the entire world last month, causing confusion and frustration everywhere. In an effort to provide some insight into how a network administrator might address the outbreak of an entirely new worm or virus, here is a recount of Ryan Smith's experiences with NIMDA.
Still undetected, the virus has now spread via the WAN link to our Corporate server. Meanwhile, the
virus is still spreading like wildfire on the IIS/home server directories.
I receive an e-mail alert that the storage space on our internal IIS/Home server was below 250 mb.
Unforunately, our users have a tendancy to dump very large database files (2gb+) onto this server
and considering the server usually only has 3-4 gb of free space, it gets filled every 6 months
or so by someone. The server still had about 240mb free on it and I assumed that was the case and
began running a Perl script to recurse the user home directories and find out who was consuming the
space. The script takes about 20 minutes to run completly so I went to grab a cup of coffee.
I received an e-mail from a user who had noticed some very strange files on the IIS/Home server, all
ending in .EML extensions. (An EML file is an Email file. It's essentially an e-mail saved to a file
and can be opened in Outlook)
Before I could look into the EML files, we received support calls from 4 different users
stating that they couldn't save any files to their home directories. I pulled up Windows Explorer
real quick and the 240mb of free space was completely gone. Still thinking it was a user's large
database file, I let my Perl script continue to run and moved a 500mb file of archived logs off
of the server and onto my local PC, just to give users some free space to work with.
I received another e-mail from a second user who also noticed the .EML files in her home directory
on the IIS/Home server. Unfortunately, she actually attached the .EML files to the e-mail. I hesitated
for a moment before opening the attachment, so instead of opening the EML file, I did a Quick View
with Windows 2000. I saw that the .EML files seemed strange. The subject was extremely long and seemed
like garbage and the body didn't contain anything. The files did have a README.EXE file attached though.
Needless to say, I was more than a little suspiscous. I immediately closed the Quick View and did a
permanent delete on the e-mail.
The full virus scan of the server turned up empty, but my File Find had found over 10,000 .EML files!
I sent a quick page off to our CIO and we kept looking into the .EML files and starting checking
some of the major antivirus sites. I had a bad feeling it was something new.
After checking Computer Associates, McAfee, Norton and Trend Micro's sites, we didn't find any new
advisories. We were thinking it might be a new strain of the APOST.A virus, simply because of the
README.EXE file that was attached. I sent a quick page again off to our CIO keeping him updated.
Russ Cooper of NTBugtraq sent an e-mail adivsory out warning of a new IIS worm propogating named
W32.nimda.amm. He only described the IIS side of the effect and didn't mention anything about
any .EML files. It seemed odd to us because, as far as we knew, our IIS side of the server was
perfectly fine. We began checking our IIS server logs to see if we could find anything else.
We sent an e-mail out to all of our users informing them of the .EML files on the network and to
NOT launch them under any circumstances. We still hadn't found anything unusual in the IIS server
logs, but that was because we weren't looking for the NIMDA/Code Red type patterns.
I ran a File Find for *.EML on the Corporate server and to my suprise discovered 468 .EML files on
that server. I had initially thought that it was isolated at least to our location, but it had
somehow spread over the WAN. I deleted all 468 .EML files, re-ran File Find and didn't find any
additional .EML files on the Corporate server.
I deleted the 11,483 .EML files on the IIS/Home server that I had discovered and started a new File
Find again.
File Find found 6,000 new .EML files on the IIS/Home server! All the files had a time stamp of
12:22 am. We were starting to feel like we were slowly losing control of the situation.
With the .EML extension of the files, we felt at the time that the point of entry for the virus was
via an e-mail message to a user who ran the README.EXE attachment.As a precaution, we decided to
shut down our Exchange Internet Mail Service to prevent any possible viruses from coming in/out
via e-mail.
Re-running the File Find on the IIS/Home Server found 16,000 .EML files now. Somehow we got an
additional 10,000 .EML files since the last File Find at 12:24!
Symantec issues their first notification. (Thank goodness for web based e-mail like Yahoo and
Hotmail!) Same general information as Russ Coopers, but still no mention of the .EML files!
Management still isn't convinced that we're suffering from the Nimda virus since we're seeing .EML
files and apparantly no one else is.
Trend Micro to the rescue! They release their first notification on it, complete with sketchy
technical details. They do however, mention the prolification of .EML files. (whew)
It took some convincing, but we got Management's approval and we're shutting down all PC's at our
location. It's never a good position to be in, but we need to ensure that we can stop the spread
of these .EML files.
Thanks to several people on the NTBugtraq list, we've got some pretty detailed preliminary instructions
on how to manually find and clean the NIMDA virus from desktop PC's. Over the next five hours,
this is done to all of the PC's. We've got twelve PC's that turn out "positive" for the NIMDA
virus and we quarantine those PC's from the network.
We check Yahoo web mail and get an e-mail that Russ Cooper of NTBugtraq sent out at 6:19 outlining
more technical information about the NIMDA virus. Full detailed instructions on the IE versions
that are effected as well as the three major attack vectors(IE, E-mail and IIS) are included.
Unfortunately, over 90% of our desktops are running Windows 9x with IE 4.x. The decision is made
to completely shut down Internet access until a patch is released for InoculateIT for NIMDA and we
get all of those PC's upgraded to IE 5.01 SP2.
Finally done, and a bit tired. Amazingly, our CIO is still here with us. Thirty minutes later, I
leave for the forty-five minutes drive home.
September 19th
7:04 am
I get to the office early after four hours of sleep and and I'm ready to go! Computer Associates
released a patch for InoculateIT overnight, but our servers couldn't upgrade to it since our
Internet is down. The other network engineer pulled it down from home and brought it in on CD-RW
and we updated the signatures of the servers.
The IIS/Home server that was initially infected is still heavily infected. For every .EML file
we delete, more take it's place. Scanning the server shows active executables running in memory
are infected with the NIMDA virus. Over 500 executables in the server are infected including
key executables for the NT operating system itself, PowerChute, WinZip, Computer Associates
Arcserve and ironically, even InoculateIT itself.
Since we're going to need to pull down critical patches, etc from the Internet, we get approval
from management to open the Net back up as well as e-mail. Considering our users at this location
are still powered down, it's not too bad. An e-mail is sent out to all the other users having them
stay off of the Net.
Fresh install of NT on the IIS/Home server begins. At least this will give us a chance to get all
the Compaq ROMPAQ's updated to the latest versions. :)
September 20th
12:38 am
Done for the day. Our IIS/Home server is rebuilt and running a restore from tape that will take
the next six hours. All of our desktop PC's have been updated to IE 5.01 SP2 and have had virus
signatures updated and are confirmed clean. None of our other servers show any signs of the NIMDA
virus still.
Restore went fine to the rebuilt IIS/Home server. Begin a full scan on all of the other servers to
see if NIMDA popped up overnight.
All of our servers are showing 100% clean now. We spend the next three hours re-assigning permissions
and finalizing last minute details on the IIS/Home Server.
Notification is given to users at our location that they can power their PC's back on and log into
the network to resume normal work operations. We all hold our breath while full scans are running
on the IIS/Home server.
Full scans show that everything is clean. Looks like we got it completely. Everything is back 100%
operational and all users are working as normal.
Summary
