[Dshield] CodeRedII - New non-variant codered worm - Analysis.

Roman Weber email at gonzo.ch
Sun Aug 5 19:19:54 GMT 2001

FYI from Incidents at Securityfocus

The following is an analysis of CodeRedII (CodeRed Version 2.0) worm.
eEye Digital Security (www.eeye.com)
Security Focus (www.securityfocus.com)

We first were contacted about this worm by the  Security Focus ARIS Incident
Analysts.  While they were monitoring various attacks from around the globe
they started to see a new attack pattern, and after a handful of packet
captures they saw there was a new worm on the loose. So they called up eEye
Digital Security to allow us to perform an analysis of this new worm.

There is in fact a completely brand new worm loose on the net right now. It
uses the same injection vector as the first CodeRed worm, however this
second worm has a completely different payload than the first worm.
Therefore this second worm is _NOT_ a variant of the first CodeRed worm.
This is an entirly new worm.

This analysis is broken up into 3 sections: 1. Infection 2. Propagation 3.

You can "follow along" in this analysis by loading the worm binary in IDA
and then following the seg locations.

This worm, like the original Code Red worm, will only exploit Windows 2000
web servers because it overwrites EIP with a jmp that is only correct under
Windows 2000. Under NT4.0 etc... that offset is different so, the process
will simply crash instead of allowing the worm to infect the system and

Analysis by Ryan Permeh (ryan at eeye.com) and Marc Maiffret (marc at eeye.com) of
eEye Digital Security (www.eeye.com). Ryan has once again generously taken
the time to comment all of the assembly code for better understanding of the

The fix that has been talked about for Code Red is still the same fix for

To check if your system has been infected or not look for the existance of
the files, c:\explorer.exe or d:\explorer.exe. Also check your IIS scripts
folder and msadc folder to see if the file root.exe exists. If it does then
you have most likely been infected with this worm. Note: An older sadmin
unicode worm also would rename cmd.exe to root.exe so you could have a bit
of cross over there.

To download this analysis and all disassembly files then goto


1st infection:

A. The first thing the worm does is setup a jump table so that it can get to
all of its needed functions.

B. The worm then proceedes to get its local IP address. This is later used
to deal with subnet masks (propagation) and to make sure that the worm does
not reinfect the local system.

C. Next, the worm gets the local System Language to see if the local system
is running Chinese (Taiwanese) or Chinese (PRC).

D. At this point the worm checks if we've executed before, and if so, then
the worm will procede to the propagation section. (See the propagation

E. Next, the worm will check to see if a CodeRedII atom has been placed
(GlobalFindAtomA). This functionality allows the worm to make sure not to
re-infect the local machine. If it sees that the atom exists then it sleeps

F. The worm will add a CodeRedII atom. This is to allow the worm the
functionality to check to see if a system has already been infected with the

G. The worm now sets its number of threads to 300 for non-Chinese systems.
If the system is Chinese then it sets it to 600.seg000:00000286

H. At this point the worm spawns a thread starting back at step A. The worm
will spawn threads according to the number set from G. Each new thread will
be a propagation thread.

I. This is where the worm calls the trojan functionality. You can find an
analysis of the trojan mechanism down below in the Trojan System section.

K. The worm then sleeps for 1 day if the local system is not Chinese, 2 days
if it is.

L. Reboot Windows.


This is used to spread the worm further.

A. Setup local IP_STORAGE variable. This is used for worm propagation
functionality and to make sure not to re-infect the local system.

B. Sleep for 64h miliseconds

C. Get local system time.  The worm checks to see if it the year is less
than 2002 or if the month is less than 10. If the date is beyond either of
those, then the worm reboots the local system. That basically limits the
worm to 10/01 for its spreading (In a perfect world.)

D. Setup SockAddr_in.  This will reference the GET_IP section.

E. Setup Socket:  This performs a Socket(), stores the handle, then makes it
a non-blocking socket (this is important for speed dealing with connect()

F. Connect to the remote host, if it returns a connect right away, goto H.

The following is how the worm generates the IP address for the next host to
connect to:

GET_IP:                                 ; CODE XREF: sub_1C4+168p

call    GET_OCTET       ; load 4th octet (this is in reverse ordwer due to
byte ordering)
mov     bh, al
call    GET_OCTET       ; get 3rd octet
mov     bl, al
shl     ebx, 10h        ; shift bx to the top of ebx
call    GET_OCTET       ; get 2nd octet
mov     bh, al
call    GET_OCTET       ; 1st
mov     bl, al
call    GEN_OCTET       ; get first octet
and     eax, 7          ; and it by 7
call    CHECK_ADDR_MASK ; ecx has eip

For each octet, generate a psuedo random byte between 1 and 254, next get a
random octet between 1 and 254 and mask it by 7
finally, use this last byte to gen a 1st octet.

most pertinent bit is CHECK_ADDR_MASK

this specifies the following:
dd 0FFFFFFFFh           ; 0 - addr masks
dd 0FFFFFF00h           ; 1
dd 0FFFFFF00h           ; 2
dd 0FFFFFF00h           ; 3
dd 0FFFFFF00h           ; 4
dd 0FFFF0000h           ; 5
dd 0FFFF0000h           ; 6
dd 0FFFF0000h           ; 7

This mask is applied to the local systems IP address, and matched to the
generated IP Address. This makes a new ip with 0,1 or 2 bytes of data with
the local ip.

For instace, the worm will 1/8th of the time generate a random IP not within
any ranges of the local IP Address.
1/2th of the time, it will stay within the same class A range of the local
IP Address
3/8th of the time, it will stay within the same class B range of the local
IP Address

Also note that if the IP the worm generates is 127.x.x.x, 224.x.x.x, or the
same as the local systems IP address then the worm will skip that IP address
and generate a new IP address to try to infect.

The way the worm generates IP addresses allows it to find more possible IIS
web servers quicker then the other CodeRed worms that have previously been
released. This new worm is also going to cause a lot more data to be zig
zaged across networks.

G. Do a select to get the handle. If no handle is returned, then goto K.

H. Set socket to Blocking. This is so select isn't required after the

I. Send a copy of the worm.

J. Do a recv. this is not actually used anywhere.

K. Close the socket and loop to A.

Trojan System

This portion of the worm is designed to dump root.exe (root.exe is cmd.exe)
into msadc and scripts, and create a trojan on the local drive.


A. Get System directory, this gets the native system directory (ie,

B. Append cmd to the system directory string (c:\winnt\system32\cmd.exe)

C. Set drive modifier to c:

D. copy cmd.exe to /scripts/root.exe (Actual path:

E. copy cmd.exe to /msadc/root.exe (Actual Path:

F.  Intitialize area for explorer.exe

G. Create Drive/explorer.exe (drive is c, then d)

H. The worm now writes out explorer.exe. There is an embedded binary within
the worm that will be written out to explorer.exe.  It has the property that
if an embedded byte is 0xFC, it geplaced by 20h 0x00 bytes instead of the
regularbyte. For more on what the trojan explorer.exe binary does then goto
the Explorer.exe Trojan section. Also the way NT works is that when a user
logs into the local system it has to load explorer.exe (desktop, task bar
etc...) however NT looks for explorer.exe first in the main drive path c:\
which means the trojan explorer.exe is going to be loaded the next time a
user logs in... therefore keeping the system trojaned over and over and

I. close explorer.exe

J. Change drive modifier to D, then the worm goes back to the code in step
D. After it is done then it goes back to step k of the infection process.

Explorer.exe Trojan
explorer.exe quick overview:

1. Get local systems windows directory.
2. Execute explorer.exe from within the local systems windows directory.
3. The worm now goes into the following loop:

set SOFTWARE\Microsoft\Windows NT\CurrentVersion\Winlogon\SFCDisable to
0FFFFFF9Dh, which basically disables system file protection.
set SYSTEM\CurrentControlSet\Services\W3SVC\Parameters\Virtual Roots\Scripts
to ,,217
set SYSTEM\CurrentControlSet\Services\W3SVC\Parameters\Virtual Roots\msadc
to ,,217
Set SYSTEM\CurrentControlSet\Services\W3SVC\Parameters\Virtual Roots\c to
Set SYSTEM\CurrentControlSet\Services\W3SVC\Parameters\Virtual Roots\d to
sleep for 10 minutes

Basically the above code creates a virtual web path (/c and /d) which maps
/c to c:\ and /d to d:\. The writer of this worm has put in this
functionality to allow for a backdoor to be placed on the system so even if
you remove the root.exe (cmd.exe prompt) from your /scripts folder an
attacker can still use the /c and /d virtual roots to compromise your
system. The attacks would basically look like:

http://IpAddress/c/inetpub/scripts/root.exe?/c+dir (if root.exe was still
there) or:
http://IpAddress/c/winnt/system32/cmd.exe?/c+dir Where dir could be any
command an attacker would want to execute.

As long as the trojan explorer.exe is running then an attacker will be able
to remotely access your server.

More information about the list mailing list