[funsec] Another WMF flaw without a Microsoft patch
H D Moore
funsecspam at digitaloffense.net
Mon Jan 9 14:27:36 CST 2006
More where that came from. The fun thing about these is that they DO apply
to Windows 96, 98, 2000-2003, Vista. You can trigger it via RTF, directly
inside IE, and anything else that loads metafiles. A fun bug you can find
in a certain WMF parsing application...:
uint_size = wmf_header.size * 2;
ptr = malloc(uint_size);
read(fd, ptr, uint_size - sizeof(wmf_header));
:-)
-HD
On Monday 09 January 2006 12:25, Richard M. Smith wrote:
> -----Original Message-----
> From: frankruder at hotmail.com [mailto:frankruder at hotmail.com]
> Sent: Saturday, January 07, 2006 12:08 AM
> To: bugtraq at securityfocus.com
> Subject: Microsoft Windows GRE WMF Format Multiple Memory Overrun
> Vulnerabilities
>
> Microsoft Windows GRE WMF Format Multiple Memory Overrun
> Vulnerabilities
>
> by cocoruder
> page:http://ruder.cdut.net
> email:frankruder_at_hotmail.com
>
> Last Update:2006.01.07
> class:design error
> Remote:yes
> local:yes
>
> Product Affected:
> Microsoft Windows XP SP2
> Microsoft Windows XP SP1
> Microsoft Windows Server 2003 SP1
> Microsoft Windows Server 2003
> Microsoft Windows ME
> Microsoft Windows 98se
> Microsoft Windows 98
> Microsoft Windows 2000SP4
>
> Vendor:
> www.microsoft.com
> Overview:
> Microsoft Windows GRE(Graphics Rendering Engine) has been discovered
> multiple memory overrun vulnerabilities while rendering WMF format
> file.Users who view the malicious WMF format file will bring a denial
> of service attack(explorer.exe will be restart).
>
> Details:
> there is 2 memory overrun vulnerabilities at least.
>
> 1.ExtCreateRegion call result in memory overrun vulnerability HRGN
> ExtCreateRegion(
> CONST XFORM *lpXform, // transformation data
> DWORD nCount, // size of region data
> CONST RGNDATA *lpRgnData // region data buffer );
>
> WMFRECORD structure:
> typedef struct _StandardMetaRecord
> {
> DWORD Size; /* Total size of the record in WORDs */
> WORD Function; /* Function number (defined in WINDOWS.H) */
> WORD Parameters[]; /* Parameter values passed to function */ }
> WMFRECORD;
>
> the Parameters is likely about:
> typedef struct _Parameters
> {
> char unknow1[0x0a];
> WORD All_PointtStruct_Num; //the total following "
> PointtStruct" number
> char unknow2[0x0a];
> char PointtStruct[]; //first "PointtStruct"
> }Parameters;
>
> "PointtStruct" structure:
> typedef struct _PointtStruct
> {
> WORD PointNum; //
> WORD Point[PointNum]; //
> char unkonow[6];
> }PointtStruct;
>
> if we set 0xff to WMFRECORD.Function,PlayMetaFileRecord will compute
> the memory which will be allocate later,as following:
>
> .text:7F00FE07 loc_7F00FE07: ; CODE XREF:
> PlayMetaFileRecord+1256j
> .text:7F00FE07 sub eax, 3
> .text:7F00FE0A jnz loc_7F022B9A ; 0xff
> .text:7F00FE10 movzx ecx, word ptr [ebx+10h] ;get
> total "PointtStruct" number
> .text:7F00FE14 mov [ebp-88h], ecx ;save
> .text:7F00FE1A test ecx, ecx
> .text:7F00FE1C jnz short loc_7F00FE2E ;jmp
> .text:7F00FE1E xor eax, eax
> .text:7F00FE20 push eax ; int
> .text:7F00FE21 push eax ; int
> .text:7F00FE22 push eax ; int
> .text:7F00FE23 push eax ; int
> .text:7F00FE24 call CreateRectRgn
> .text:7F00FE29 jmp loc_7F010494
> .text:7F00FE2E ;
> 哪哪哪哪哪哪哪哪哪
>;
> 738;哪哪哪哪哪哪哪哪
>1738;
> 哪哪哪哪哪哪哪哪哪
>;
> 738;哪哪哪哪哪哪哪哪?
> .text:7F00FE2E
> .text:7F00FE2E loc_7F00FE2E: ; CODE XREF:
> PlayMetaFileRecord+C15j
> .text:7F00FE2E xor edi, edi
> .text:7F00FE30 mov [ebp-8Ch], edi
> .text:7F00FE36 lea eax, [ebx+1Ch] ;get first
> "PointtStruct" address
> .text:7F00FE39 mov [ebp-90h], eax ;save
> .text:7F00FE3F and [ebp-94h], edi
> .text:7F00FE45 and [ebp-98h], edi
> .text:7F00FE4B test ecx, ecx
> .text:7F00FE4D jbe short loc_7F00FE8C
> .text:7F00FE4F
> .text:7F00FE4F loc_7F00FE4F: ; CODE XREF:
> PlayMetaFileRecord+C83j
> .text:7F00FE4F movzx ecx, word ptr [eax] ;get
> PointNum,here will trigger memory access error**
> .text:7F00FE52 mov edx, ecx
> .text:7F00FE54 shr edx, 1 ;PointNum/2
> .text:7F00FE56 add edx, edi ;sum
> PointNum/2
> .text:7F00FE58 cmp edx, edi
> .text:7F00FE5A jb loc_7F0106D5
> .text:7F00FE60 mov edi, edx
> .text:7F00FE62 mov [ebp-8Ch], edi
> .text:7F00FE68 cmp ecx, 7FFFFFFBh
> .text:7F00FE6E jnb loc_7F0106D5
> .text:7F00FE74 lea eax, [eax+ecx*2+8]
> .text:7F00FE78 inc dword ptr [ebp-98h] ;counter+1
> .text:7F00FE7E mov ecx, [ebp-98h]
> .text:7F00FE84 cmp ecx, [ebp-88h] ;cmp total
> "PointtStruct" number
> .text:7F00FE8A jb short loc_7F00FE4F ;traverse
> all "PointtStruct"
> .text:7F00FE8C
> .text:7F00FE8C loc_7F00FE8C: ; CODE XREF:
> PlayMetaFileRecord+C46j
> .text:7F00FE8C ;
> PlayMetaFileRecord+14D8j
> .text:7F00FE8C cmp dword ptr [ebp-94h], 0
> .text:7F00FE93 jnz loc_7F022BA2
> .text:7F00FE99 cmp edi, 0FFFFFFDh
> .text:7F00FE9F jnb loc_7F022BA2
> .text:7F00FEA5 mov eax, [ebp-8Ch]
> .text:7F00FEAB add eax, 2
> .text:7F00FEAE shl eax, 4
> .text:7F00FEB1 mov [ebp-9Ch], eax
> .text:7F00FEB7 push eax ; uBytes
> .text:7F00FEB8 push 0 ; uFlags
> .text:7F00FEBA call ds:LocalAlloc ;will
> allocate memory normally.
> .text:7F00FEC0 mov edi, eax
>
> so when we set
> "Parameters.All_PointtStruct_Num","PointtStruct.PointNum" big
> enough,but not many enough "PointtStruct" structure,will trigger a
> memory access error on 7F00FE4F.
>
>
> 2.ExtEscape POSTSCRIPT_INJECTION result in memory overrun vulnerability
> int ExtEscape(
> hdc, // handle to DC (HDC)
> POSTSCRIPT_INJECTION, // nEscape param of ExtEscape
> cbInput, // size of input buffer
> lpszInData, // pointer to input (PSINJECTDATA *)
> 0, // cbOutput param of ExtEscape
> NULL // lpszOutData param of ExtEscape
> );
>
> we can control cbInput and szInData,so when we set cbInput big
> enough(like 0xffff),but set szInData very small,will trigger a memory
> access error(7F027358) likely.
>
> .text:7F027312 loc_7F027312: ; CODE XREF:
> ExtEscape+11Ej
> .text:7F027312 ;
> ExtEscape+12Aj .text:7F027312 test byte ptr
> [ecx+4], 40h
> .text:7F027316 jnz loc_7F017CEC
> .text:7F02731C mov ebx, [ebp+arg_8] ;we can
> cotrol this:cbSize
> .text:7F02731F add ebx, 1Ah ;cbSize+0x1a
> .text:7F027322 and ebx, 0FFFFFFFCh
> ;cbSize+0x1a-4
> .text:7F027325 mov eax, large fs:18h
> .text:7F02732B mov eax, [eax+30h]
> .text:7F02732E push ebx
> .text:7F02732F push 0
> .text:7F027331 push dword ptr [eax+18h]
> .text:7F027334 call ds:RtlAllocateHeap ;allocate
> memory size=cbSize+0x16
> .text:7F02733A test eax, eax
> .text:7F02733C jz short loc_7F027385
> .text:7F02733E mov ecx, [ebp+arg_4]
> .text:7F027341 mov [eax+0Ch], ecx
> .text:7F027344 mov ecx, [ebp+arg_8]
> .text:7F027347 mov [eax+10h], ecx
> .text:7F02734A mov edx, ecx
> .text:7F02734C shr ecx, 2
> .text:7F02734F sub ebx, 8
> .text:7F027352 mov [eax+8], ebx
> .text:7F027355 lea edi, [eax+14h]
> .text:7F027358 rep movsd ;copy,here
> will trigger a memory error likely**
> .text:7F02735A mov ecx, edx
> .text:7F02735C and ecx, 3
> .text:7F02735F rep movsb
> .text:7F027361 mov ecx, [ebp-4]
> .text:7F027364 mov edi, [ebp+arg_14]
> .text:7F027367 lea edx, [ecx+48h]
> .text:7F02736A mov esi, [edx+4]
> .text:7F02736D mov [eax+4], esi
>
> Exploit:
> first vul can test like this:
>
> ...05020000FF0044444444444444444444FFFF33333333333333333333FFFF41424344
>... second vul can test like this:
> ...0502000026001610FFFF4444444444444444...
> my blog can not be opened now:p,i will upload the exploit files to
> xfocus(www.xfocus.net) later,have fun:)
>
> Solution:
> Microsoft has not develop the patch,please unregister the Windows
> Picture and Fax Viewer (Shimgvw.dll)(see MS06-001).
>
> Thanks:
> thanks all my friends,Happy Weekday:)
>
> -EOF-
>
> _______________________________________________
> Fun and Misc security discussion for OT posts.
> https://linuxbox.org/cgi-bin/mailman/listinfo/funsec
> Note: funsec is a public and open mailing list.
More information about the funsec
mailing list