This chapter focuses on the technical execution of IP network scanning. After undertaking initial stealthy reconnaissance to identify IP address spaces of interest, network scanning is an intrusive and aggressive process used to identify accessible hosts and their network services. The rationale behind IP network scanning is to gain insight into the following elements of a given network:
ICMP message types that generate responses from target hosts
Accessible TCP and UDP network services running on the target hosts
Operating platforms of target hosts and their configurations
Areas of vulnerability within target host IP stack implementations (including sequence number predictability for TCP spoofing and session hijacking)
Configuration of filtering and security systems (including firewalls, border routers, switches, and IDS/IPS mechanisms)
Performing both network scanning and reconnaissance tasks paints a clear picture of the network topology and its security features. Before penetrating the target network, specific network service probing is undertaken to enumerate vulnerabilities and weaknesses, covered in later chapters of this book.
Internet Control Message Protocol (ICMP) probes can be used to identify potentially weak and poorly protected networks and hosts. ICMP is a short messaging protocol, used by systems administrators for continuity testing of networks in particular (using tools such as ping and traceroute). From a network scanning perspective, the following types of ICMP messages are useful:
The ICMP information request message was intended to support self-configuring systems such as diskless workstations at boot time to allow them to discover their network addresses. Protocols such as RARP, BOOTP, or DHCP achieve this more robustly, so type 15 messages are rarely used.
An address mask request message reveals the subnet mask used by the target host. This information is useful when mapping networks and identifying the size of subnets and network spaces used by organizations.
Firewalls of security-conscious organizations often blanket-filter inbound ICMP messages, and so ICMP probing isn’t effective; however, ICMP isn’t filtered in most cases, as these messages are useful during network troubleshooting.
Send ICMP Nasty Garbage (SING) is a
command-line utility that sends customizable ICMP probes. The main
purpose of the tool is to replace the
ping command with certain enhancements,
including the ability to transmit and receive spoofed packets, send
MAC-spoofed packets, and support the transmission of many other
message types, including ICMP address mask, timestamp, and
information requests, as well as router solicitation and router
SING is available from http://sourceforge.net/projects/sing. Examples using SING to launch ICMP echo, timestamp, and address mask requests follow. In these examples, I direct probes at broadcast addresses and individual hosts.
sing -echo 192.168.0.255SINGing to 192.168.0.255 (192.168.0.255): 16 data bytes 16 bytes from 192.168.0.1: seq=0 ttl=64 TOS=0 time=0.230 ms 16 bytes from 192.168.0.155: seq=0 ttl=64 TOS=0 time=2.267 ms 16 bytes from 192.168.0.126: seq=0 ttl=64 TOS=0 time=2.491 ms 16 bytes from 192.168.0.50: seq=0 ttl=64 TOS=0 time=2.202 ms 16 bytes from 192.168.0.89: seq=0 ttl=64 TOS=0 time=1.572 ms
sing -tstamp 192.168.0.50SINGing to 192.168.0.50 (192.168.0.50): 20 data bytes 20 bytes from 192.168.0.50: seq=0 ttl=128 TOS=0 diff=327372878 20 bytes from 192.168.0.50: seq=1 ttl=128 TOS=0 diff=1938181226* 20 bytes from 192.168.0.50: seq=2 ttl=128 TOS=0 diff=1552566402* 20 bytes from 192.168.0.50: seq=3 ttl=128 TOS=0 diff=1183728794*
sing -mask 192.168.0.25SINGing to 192.168.0.25 (192.168.0.25): 12 data bytes 12 bytes from 192.168.0.25: seq=0 ttl=236 mask=255.255.255.0 12 bytes from 192.168.0.25: seq=1 ttl=236 mask=255.255.255.0 12 bytes from 192.168.0.25: seq=2 ttl=236 mask=255.255.255.0 12 bytes from 192.168.0.25: seq=3 ttl=236 mask=255.255.255.0
There are a handful of other ICMP message types that have other security implications, such as ICMP type 5 redirect messages sent by routers, which allow for traffic redirection. These messages aren’t related to network scanning, and so they are not detailed here. For details of traffic redirection using ICMP, including exploit code, please see Yuri Volobuev’s BugTraq post at http://seclists.org/bugtraq/1997/sep/0057.html.
Nmap (http://insecure.org/nmap/) can
perform ICMP ping sweep scans of target IP blocks easily. Many
hardened networks will blanket-filter inbound ICMP messages at
border routers or firewalls, so sweeping in this fashion isn’t
effective in some cases. Nmap can be run from a Unix-based or
Windows command prompt to perform an ICMP ping sweep against
192.168.0.0/24, as shown in Example 4-1.
nmap -sP -PI 192.168.0.0/24Starting Nmap 4.10 ( http://www.insecure.org/nmap/ ) at 2007-04-01 20:39 UTC Host 192.168.0.0 seems to be a subnet broadcast address (2 extra pings). Host 192.168.0.1 appears to be up. Host 192.168.0.25 appears to be up. Host 192.168.0.32 appears to be up. Host 192.168.0.50 appears to be up. Host 192.168.0.65 appears to be up. Host 192.168.0.102 appears to be up. Host 192.168.0.110 appears to be up. Host 192.168.0.155 appears to be up. Host 192.168.0.255 seems to be a subnet broadcast address (2 extra pings). Nmap finished: 256 IP addresses (8 hosts up) scanned in 17.329 seconds
-sP ping sweep
flag within Nmap doesn’t just perform an ICMP echo request to each
IP address; it also sends TCP ACK and SYN probe packets to port 80
of each host. In Example 4-10, Nmap is run with
-sP flag to specify that
we’re sending only ICMP echo requests. Overall, using the standard
-sP flag is often more
effective because it identifies web servers that may not respond
to ICMP probes; however, in some environments it is beneficial to
use more specific probe types.
ICMPScan is a bulk scanner that sends type 8, 13, 15, and 17
ICMP messages, derived from Nmap and available from http://www.bindshell.net/tools/icmpscan. The tool is
very useful in that it can process inbound responses by placing the
network interface into promiscuous mode, thereby identifying
internal IP addresses and machines that respond from probes sent to
subnet network and broadcast addresses. Example 4-2 shows ICMPScan being run against an
internal network block. Because ICMP is a connectionless protocol,
it is best practice to resend each probe (using
-r 1) and set the timeout to 500
-t 500). We
also set the tool to listen in promiscuous mode for unsolicited
responses (using the
icmpscanUsage: icmpscan [ options ] target [...] -i <interface> Specify interface. -c Enable promiscuous mode. -A <address> Specify source address of generated packets. -t <timeout> Specify timeout for probe response. -r <retries> Retries per probe. -f <filename> Read targets from the specified file. -E, -P ICMP Echo Probe -T, -S Timestamp -N, -M Netmask -I Info -R Router solicitation -h Display usage information -v Increase verbosity -B Enable debugging output. -n Numeric output (do not resolve hostnames) $
icmpscan -c -t 500 -r 1 192.168.1.0/24192.168.1.0: Echo (From 192.168.1.17!) 192.168.1.0: Address Mask [255.255.255.0] (From 192.168.1.17!) 192.168.1.7: Echo 192.168.1.7: Timestamp [0x03ab2db0, 0x02d4c507, 0x02d4c507] 192.168.1.7: Address Mask [255.255.255.0] 192.168.1.8: Echo 192.168.1.8: Address Mask [255.255.255.0]
Nmap identifies subnet network and broadcast addresses by
counting the number of ICMP echo replies for each IP address during an
ICMP ping sweep. Such addresses respond with multiple replies,
providing insight into the target network and its segmentation. In
Example 4-3 we use Nmap to
enumerate subnet network and broadcast addresses in use for a given
nmap -sP 18.104.22.168/26Starting Nmap 4.10 ( http://www.insecure.org/nmap/ ) at 2007-04-01 20:39 UTC Host 22.214.171.124 seems to be a subnet broadcast address (returned 1 extra pings). Host pipex-gw.abc.co.uk (126.96.36.199) appears to be up. Host mail.abc.co.uk (188.8.131.52) appears to be up. Host 184.108.40.206 appears to be up. Host intranet.abc.co.uk (220.127.116.11) appears to be up. Host 18.104.22.168 appears to be up. Host 22.214.171.124 appears to be up. Host 126.96.36.199 seems to be a subnet broadcast address (returned 1 extra pings). Host 188.8.131.52 seems to be a subnet broadcast address (returned 1 extra pings). Host pipex-gw.smallco.net (184.108.40.206) appears to be up. Host mail.smallco.net (220.127.116.11) appears to be up. Host 18.104.22.168 seems to be a subnet broadcast address (returned 1 extra pings). Host 22.214.171.124 seems to be a subnet broadcast address (returned 1 extra pings). Host pipex-gw.example.org (126.96.36.199) appears to be up. Host gatekeeper.example.org (188.8.131.52) appears to be up. Host 184.108.40.206 appears to be up. Host 220.127.116.11 seems to be a subnet broadcast address (returned 1 extra pings).
This scan has identified six subnets within the
18.104.22.168/26 network, as follows:
An unused or filtered block from
22.214.171.124 (14 usable
abc.co.uk block from
126.96.36.199 (14 usable
188.8.131.52 (2 usable
An unused or filtered block from
184.108.40.206 (2 usable
220.127.116.11 (6 usable
An unused or filtered block from
18.104.22.168 (14 usable
Useful details about subnet network and broadcast addresses and CIDR slash notation can be found at http://en.wikipedia.org/wiki/classless_inter-domain_routing. An online IP calculator is also available at http://jodies.de/ipcalc.
In some cases, it is possible to gather internal IP address information by analyzing ICMP responses from an ICMP ping sweep. Upon sending ICMP echo requests to publicly accessible IP addresses, firewalls often use Network Address Translation (NAT) or similar IP masquerading to forward the packets on to internal addresses, which then respond to the probes. Other scenarios include poor routing configuration on routers that are probed using ICMP, where they respond to the probes from a different interface.
Stateful inspection mechanisms and sniffers can be used to monitor for ICMP responses from internal IP addresses in relation to your original probes. Tools such as Nmap and SING don’t identify these responses from private addresses, as low-level stateful analysis of the traffic flowing into and out of a network is required. A quick and simple example of this behavior can be seen in the ISS BlackICE personal firewall event log in Figure 4-1 as a simple ICMP ping sweep is performed.
This figure shows that BlackICE has identified four unsolicited
ICMP echo replies from private addresses (within the
172.16.0.0/12 space in this case, but they
are often within
ICMPScan supports this type of internal IP address discovery when in promiscuous mode. It is beneficial to run a network sniffer such as Ethereal or tcpdump during testing to pick up on unsolicited ICMP responses, including “ICMP TTL exceeded” (type 11 code 0) messages, indicating a routing loop, and “ICMP administratively prohibited” (type 3 code 13) messages, indicating an ACL in use on a router or firewall.
Ofir Arkin’s Xprobe2 utility performs OS fingerprinting by primarily analyzing responses to ICMP probes. See the Sys-Security Group web site (http://www.sys-security.com) for further details, including white papers and presentations that describe the Xprobe2 fingerprinting technology and approach. Example 4-4 shows Xprobe2 being used to fingerprint a remote host.
xprobe2 -v 192.168.0.174Xprobe2 v.0.3 Copyright (c) 2002-2005 email@example.com, firstname.lastname@example.org, email@example.com [+] Target is 192.168.0.174 [+] Loading modules. [+] Following modules are loaded: [x]  ping:icmp_ping - ICMP echo discovery module [x]  ping:tcp_ping - TCP-based ping discovery module [x]  ping:udp_ping - UDP-based ping discovery module [x]  infogather:ttl_calc - TCP and UDP based TTL distance calculation [x]  infogather:portscan - TCP and UDP PortScanner [x]  fingerprint:icmp_echo - ICMP Echo request fingerprinting module [x]  fingerprint:icmp_tstamp - ICMP Timestamp request fingerprinting module [x]  fingerprint:icmp_amask - ICMP Address mask request fingerprinting module [x]  fingerprint:icmp_port_unreach - ICMP port unreachable fingerprinting module [x]  fingerprint:tcp_hshake - TCP Handshake fingerprinting module [x]  fingerprint:tcp_rst - TCP RST fingerprinting module [x]  fingerprint:smb - SMB fingerprinting module [x]  fingerprint:snmp - SNMPv2c fingerprinting module [+] 13 modules registered [+] Initializing scan engine [+] Running scan engine [+] Host: 192.168.0.174 is up (Guess probability: 100%) [+] Target: 192.168.0.174 is alive. Round-Trip Time: 0.00015 sec [+] Selected safe Round-Trip Time value is: 0.00030 sec [+] Primary guess: [+] Host 192.168.0.174 Running OS: "Sun Solaris 5 (SunOS 2.5)" (Guess probability: 100%) [+] Other guesses: [+] Host 192.168.0.174 Running OS: "Sun Solaris 6 (SunOS 2.6)" (Guess probability: 100%) [+] Host 192.168.0.174 Running OS: "Sun Solaris 7 (SunOS 2.7)" (Guess probability: 100%) [+] Host 192.168.0.174 Running OS: "Sun Solaris 8 (SunOS 2.8)" (Guess probability: 100%) [+] Host 192.168.0.174 Running OS: "Sun Solaris 9 (SunOS 2.9)" (Guess probability: 100%) [+] Host 192.168.0.174 Running OS: "Mac OS 9.2.x" (Guess probability: 95%) [+] Host 192.168.0.174 Running OS: "HPUX B.11.0 x" (Guess probability: 95%) [+] Host 192.168.0.174 Running OS: "Mac OS X 10.1.5" (Guess probability: 87%) [+] Host 192.168.0.174 Running OS: "FreeBSD 4.3" (Guess probability: 87%) [+] Host 192.168.0.174 Running OS: "FreeBSD 4.2" (Guess probability: 87%)
Proxy bounce scanning
Sniffer-based spoofed scanning
What follows is a technical breakdown for each TCP port scanning type, along with details of Windows- and Unix-based tools that can perform scanning.
Standard scanning methods, such as vanilla and half-open SYN scanning, are extremely simple direct techniques used to accurately identify accessible TCP ports and services. These scanning methods are reliable but are easily logged and identified.
connect( ) port
scanning is the simplest type of probe to launch. There is no
stealth whatsoever involved in this form of scanning, as a full
TCP/IP connection is established with each port of the target
TCP/IP robustness means that
) port scanning is an accurate way to determine which TCP
services are accessible on a given host. However, due to the way
that a full three-way handshake is performed, an aggressive
connect( ) scan could antagonize or break
poorly written network services. Figure 4-2 and Figure 4-3 show the various
TCP packets involved and their flags.
In Figure 4-2, the attacker first sends a SYN probe packet to the port he wishes to test. Upon receiving a packet from the port with the SYN and ACK flags set, he knows that the port is open. The attacker completes the three-way handshake by sending an ACK packet back.
If, however, the target port is closed, the attacker receives an RST/ACK packet directly back, as shown in Figure 4-3.
Nmap can perform a TCP
) port scan using the
-sT flag. A benefit of this scanning
type is that superuser root access is not required, as raw network
sockets are not used. Other very simple scanners exist, including
pscan.c, which is available
as source code from many sites including Packet Storm (http://www.packetstormsecurity.org).
When performing a full assessment exercise, every TCP port from 0 to 65535 should be checked. For speed reasons, port scanners such as Nmap have internal lists of only some 1,500 common ports to check; thus, they often miss all kinds of interesting services that can be found on high ports, for example, Check Point SVN web services on TCP port 18264.
Usually, a three-way handshake is initiated to synchronize a connection between two hosts; the client sends a SYN packet to the server, which responds with SYN and ACK if the port is open, and the client then sends an ACK to complete the handshake.
In the case of half-open SYN port scanning, when a port is found to be listening, an RST packet is sent as the third part of the handshake. Sending an RST packet in this way abruptly resets the TCP connection, and because you have not completed the three-way handshake, the connection attempt often isn’t logged on the target host.
Most network-based Intrusion Detection Systems (IDSs) and other security mechanisms, such as portsentry, can detect half-open SYN port scanning attempts. In cases where stealth is required, other techniques are recommended, such as FIN or TTL-based scanning, and fragmenting outbound packets to avoid detection.
Figure 4-4 shows that when a closed port is found, a RST/ACK packet is received, and nothing happens (as before in Figure 4-3). Benefits of half-open scanning include speed and efficiency (fewer packets are sent and received), and the fact that the connection isn’t established, which can bypass some logging mechanisms.
In Figure 4-5, a SYN
probe packet is sent to the target port and a SYN/ACK packet is
received indicating that the port is open. Normally at this stage, a
connect( ) scanner sends an ACK
packet to establish the connection, but this is half-open scanning,
so instead, an RST packet is sent to tear down the
Nowadays, all IDS and personal firewall systems can identify SYN port scans (although they often mislabel them as SYN flood attacks due to the number of probe packets). SYN scanning is fast and reliable, although it requires raw access to network sockets and therefore requires privileged access to Unix and Windows hosts.
Nmap can perform a SYN port scan using the
-sS flag. Another SYN port scanner worth
mentioning is Scanrand, a component of the Paketto Keiretsu suite.
Paketto Keiretsu contains a number of useful networking utilities
that are available at http://www.doxpara.com/read.php/code/paketto.html.
For Windows, Foundstone’s SuperScan is an excellent port scanning
utility with good functionality, including banner grabbing.
SuperScan is available from
-T flag can be used
within Nmap to change the scanning timing policy. Networks
protected by commercial firewalls (NetScreen, WatchGuard, and
Check Point in particular) will often drop SYN probes if Nmap is
sending the packets out too quickly because Nmap’s actions
resemble a SYN flood Denial of Service (DoS) attack. I have
found that by setting the timing policy to
-T Sneaky, it’s often possible to
glean accurate results against hosts protected by firewalls with
SYN flood protection enabled.
Scanrand is well designed, with distinct SYN probing and background listening components that allow for very fast scanning. Inverse SYN cookies (using SHA1) tag outgoing probe packets, so that false positive results become nonexistent, as the listening component only registers responses with the correct SYN cookies. Example 4-5 shows Scanrand identifying open ports on a local network in less than one second.
scanrand 10.0.1.1-254:quickUP: 10.0.1.38:80  0.003s UP: 10.0.1.110:443  0.017s UP: 10.0.1.254:443  0.021s UP: 10.0.1.57:445  0.024s UP: 10.0.1.59:445  0.024s UP: 10.0.1.38:22  0.047s UP: 10.0.1.110:22  0.058s UP: 10.0.1.110:23  0.058s UP: 10.0.1.254:22  0.077s UP: 10.0.1.254:23  0.077s UP: 10.0.1.25:135  0.088s UP: 10.0.1.57:135  0.089s UP: 10.0.1.59:135  0.090s UP: 10.0.1.25:139  0.097s UP: 10.0.1.27:139  0.098s UP: 10.0.1.57:139  0.099s UP: 10.0.1.59:139  0.099s UP: 10.0.1.38:111  0.127s UP: 10.0.1.57:1025  0.147s UP: 10.0.1.59:1025  0.147s UP: 10.0.1.57:5000  0.156s UP: 10.0.1.59:5000  0.157s UP: 10.0.1.53:111  0.182s
Due to the way Scanrand sends a deluge of SYN probes and then listens for positive SYN/ACK responses, the order in which the open ports are displayed will look a little odd. On the positive side, Scanrand is much faster than bulkier scanners, such as Nmap.
Unicornscan (http://www.unicornscan.org) is another tool that performs fast half-open scanning. It has some unique and very useful features, and it is recommended for advanced users.
Stealth scanning methods take advantage of idiosyncrasies in certain TCP/IP stack implementations. Such techniques aren’t effective at accurately mapping the open ports of some operating systems, but they do provide a degree of stealth when susceptible platforms are found.
Security mechanisms such as firewalls and IDS usually detect SYN packets being sent to sensitive ports of target hosts. To avoid this detection, we can send probe packets with different TCP flags set.
Using malformed TCP flags to probe a target is known as an inverted technique because responses are sent back only by closed ports. RFC 793 states that if a port is closed on a host, an RST/ACK packet should be sent to reset the connection. To take advantage of this feature, attackers send TCP probe packets with various TCP flags set.
A TCP probe packet is sent to each port of the target host. Three types of probe packet flag configurations are normally used:
The RFC standard states that if no response is seen from the target port, either the port is open or the server is down. This scanning method isn’t necessarily the most accurate, but it is stealthy; it sends garbage that usually won’t be picked up to each port.
For all closed ports on the target host, RST/ACK packets are received. However, some operating platforms (such as those in the Microsoft Windows family) disregard the RFC 793 standard, so no RST/ACK response is seen when an attempt is made to connect to a closed port. Hence, this technique is effective against some Unix-based platforms.
Vscan is another Windows tool you can use to perform inverse TCP flag scanning. The utility doesn’t require installation of WinPcap network drivers; instead it uses raw sockets within Winsock 2 (present in Windows itself). Vscan is available from http://examples.oreilly.com/networksa/tools/vscan.zip.
A stealthy technique documented by Uriel Maimon in Phrack Magazine, issue 49, is that of identifying open TCP ports by sending ACK probe packets and analyzing the header information of the RST packets received from the target host. This technique exploits vulnerabilities within the BSD-derived TCP/IP stack and is therefore only effective against certain operating systems and platforms. There are two main ACK scanning techniques:
Analysis of the time-to-live (TTL) field of received packets
Analysis of the WINDOW field of received packets
These techniques can also check filtering systems and complicated networks to understand the processes packets go through on the target network. For example, the TTL value can be used as a marker of how many systems the packet has hopped through. The Firewalk filter assessment tool works in a similar fashion, available from http://www.packetfactory.net/projects/firewalk.
To analyze the TTL field data of received RST packets, an attacker first sends thousands of crafted ACK packets to different TCP ports, as shown in Figure 4-8.
Here is a log of the first four RST packets received using Hping2:
1: host 192.168.0.12 port 20: F:RST -> ttl: 70 win: 0 2: host 192.168.0.12 port 21: F:RST -> ttl: 70 win: 0 3: host 192.168.0.12 port 22: F:RST -> ttl: 40 win: 0 4: host 192.168.0.12 port 23: F:RST -> ttl: 70 win: 0
By analyzing the TTL value of each packet, an attacker can easily see that the value returned by port 22 is 40, whereas the other ports return a value of 70. This suggests that port 22 is open on the target host because the TTL value returned is smaller than the TTL boundary value of 64.
To analyze the WINDOW field data of received RST packets, an attacker sends thousands of the same crafted ACK packets to different TCP ports (as shown in Figure 4-8). Here is a log of the first four RST packets received, again using Hping2:
1: host 192.168.0.20 port 20: F:RST -> ttl: 64 win: 0 2: host 192.168.0.20 port 21: F:RST -> ttl: 64 win: 0 3: host 192.168.0.20 port 22: F:RST -> ttl: 64 win: 512 4: host 192.168.0.20 port 23: F:RST -> ttl: 64 win: 0
Notice that the TTL value for each packet is 64, meaning that TTL analysis of the packets isn’t effective in identifying open ports on this host. However, by analyzing the WINDOW values, the attacker finds that the third packet has a nonzero value, indicating an open port.
The advantage of using ACK flag probe scanning is that detection is difficult (for both IDS and host-based systems, such as personal firewalls). The disadvantage is that this scanning type relies on TCP/IP stack implementation bugs, which are prominent in BSD-derived systems but not in many other modern platforms.
Nmap supports ACK flag probe scanning, with the
-sW flags to analyze the TTL and WINDOW
values, respectively. See the Nmap manual page for more detailed
Hping2 can also sample TTL and WINDOW values, but this can prove highly time-consuming in most cases. The tool is more useful for analyzing low-level responses, as opposed to port scanning in this fashion. Hping2 is available from http://www.hping.org.
Third-party port scanning methods allow for probes to be effectively bounced through vulnerable servers to hide the true source of the network scanning. An additional benefit of using a third-party technique in this way is that insight into firewall configuration can be gained by potentially bouncing scans through trusted hosts that are vulnerable.
Hosts running outdated FTP services can relay numerous TCP
attacks, including port scanning. There is a flaw in the way many
FTP servers handle connections using the
PORT command (see RFC 959 or technical
description) that allows data to be sent to user-specified hosts and
ports. In their default configurations, the FTP services running on
the following older Unix-based platforms are affected:
FreeBSD 2.1.7 and earlier
HP-UX 10.10 and earlier
Solaris 2.6 and earlier
SunOS 4.1.4 and earlier
SCO OpenServer 5.0.4 and earlier
SCO UnixWare 2.1 and earlier
IBM AIX 4.3 and earlier
Caldera Linux 1.2 and earlier
Red Hat Linux 4.2 and earlier
Slackware 3.3 and earlier
Any Linux distribution running WU-FTP 2.4.2-BETA-16 or earlier
The FTP bounce attack can have a far more devastating effect
if a writable directory exists because a series of commands or other
data can be entered into a file and then relayed via the
PORT command to a specified port of a
target host. For example, someone can upload a spam email message to
a vulnerable FTP server and then send this email message to the SMTP
port of a target mail server. Figure 4-9 shows the parties involved in
FTP bounce scanning.
The following occurs when performing an FTP bounce scan:
The attacker connects to the FTP control port (TCP port 21) of the vulnerable FTP server that she is going to bounce her attack through and enters passive mode, forcing the FTP server to send data to a specific port of a specific host:
QUOTE PASV227 Entering Passive Mode (64,12,168,246,56,185).
PORT command is
issued, with an argument passed to the FTP service telling it to
attempt a connection to a specific TCP port on the target
server; for example, TCP port 23 of
PORT 144,51,17,230,0,23200 PORT command successful.
After issuing the
LIST command is
sent. The FTP server then attempts to create a connection with
the target host defined in the
PORT command issued previously:
LIST150 Opening ASCII mode data connection for file list 226 Transfer complete.
226 response is
seen, then the port on the target host is open. If, however, a
425 response is seen, the
connection has been refused:
LIST425 Can't build data connection: Connection refused
nmap -P0 -b username:password@ftp-server:port <target host>
-P0 flag must be used
to suppress pinging of the target host, as it may not be
accessible from your location (e.g., if you are bouncing through a
multihomed FTP server). Also, you may not want your source IP
address to appear in logs at the target site.
Attackers bounce TCP attacks through open proxy servers. Depending on the level of poor configuration, the server will sometimes allow a full-blown TCP port scan to be relayed. Using proxy servers to perform bounce port scanning in this fashion is often time-consuming, so many attackers prefer to abuse open proxy servers more efficiently by bouncing actual attacks through to target networks.
ppscan.c, a publicly available Unix-based tool to bounce port scans, can be found in source form at:
An innovative half-open SYN TCP port scanning method was realized when jsbach published his Unix-based scanner, spoofscan, in 1998. The spoofscan tool is run as root on a given host to perform a stealthy port scan. The key feature that makes this scanner so innovative is that it places the host network card into promiscuous mode and then sniffs for responses on the local network segment.
The following unique benefits are immediately realized when using a sniffer-based spoofing port scanner:
If you have administrator access to a machine on the same physical network segment as the target host or a firewall protecting a target host, you can spoof TCP probes from other IP addresses to identify trusted hosts and to gain insight into the firewall policy (by spoofing scans from trusted office hosts, for example). Accurate results will be retrieved because of the background sniffing process, which monitors the local network segment for responses to your spoofed probes.
If you have access to a large shared network segment, you can spoof scans from hosts you don’t have access to or that don’t exist (such as unused IP addresses within your local network segment), to effectively port scan remote networks in a distributed and stealthy fashion.
The beauty of this method is that the attacker is abusing his access to the local network segment. Such techniques can even be carried out to good effect in switched network environments using ARP redirect spoofing and other techniques. spoofscan is available at http://examples.oreilly.com/networksa/tools/spoofscan.c.
IP ID header scanning (also known as idle or dumb scanning) is an obscure scanning technique that involves abusing implementation peculiarities within the TCP/IP stack of most operating systems. Three hosts are involved:
The host from which the scan is launched
The target host that will be scanned
A zombie or idle host, which is an Internet-based server that is queried with spoofed port scanning against the target host to identify open ports from the perspective of the zombie host
IP ID header scanning is extraordinarily stealthy due to its blind nature. Determined attackers will often use this type of scan to map out IP-based trust relationships between machines, such as firewalls and VPN gateways.
The listing returned by the scan shows open ports from the perspective of the zombie host, so you can try scanning a target using various zombies you think might be trusted (such as hosts at remote offices or DMZ machines). Figure 4-10 depicts the process undertaken during an IP ID header scan.
Hping2 was originally used in a manual fashion to perform such low-level TCP scanning, which was time-consuming and tricky to undertake against an entire network of hosts. A white paper that fully discusses using the tool to perform IP ID header scanning by hand is available from http://www.kyuzz.org/antirez/papers/dumbscan.html.
Nmap supports such IP ID header scanning with the option:
-sI <zombie host[:probe port]>
By default, Nmap uses port 80 to perform this scanning through
the zombie host. Example 4-6 shows how Nmap is
used to scan
nmap -P0 -sI 192.168.0.155 192.168.0.50Starting Nmap 4.10 ( http://www.insecure.org/nmap/ ) at 2007-04-01 23:24 UTC Idlescan using zombie 192.168.0.155; Class: Incremental Interesting ports on (192.168.0.50): (The 1582 ports scanned but not shown below are in state: closed) Port State Service 25/tcp open smtp 53/tcp open domain 80/tcp open http 88/tcp open kerberos-sec 135/tcp open loc-srv 139/tcp open netbios-ssn 389/tcp open ldap 443/tcp open https 445/tcp open microsoft-ds 464/tcp open kpasswd5 593/tcp open http-rpc-epmap 636/tcp open ldapssl 1026/tcp open LSA-or-nterm 1029/tcp open ms-lsa 1033/tcp open netinfo 3268/tcp open globalcatLDAP 3269/tcp open globalcatLDAPssl 3372/tcp open msdtc 3389/tcp open ms-term-serv
If Nmap is run without the
-P0 flag when performing third-party
scanning, the source IP address of the attacker’s host performs
ICMP and TCP pinging of the target hosts before starting to scan;
this can appear in firewall and IDS audit logs of
Vscan is another Windows tool you can use to perform IP ID header scanning. As discussed earlier, the utility doesn’t require installation of WinPcap network drivers; instead it uses raw sockets within Winsock 2 (present in Windows itself). Vscan is available from http://examples.oreilly.com/networksa/tools/vscan.zip.
Figure 4-11 shows Vscan in use, along with its options and functionality.
Send UDP probe packets to all 65535 UDP ports, then wait for “ICMP destination port unreachable” messages to identify UDP ports that aren’t accessible.
Use specific UDP service clients (such as snmpwalk, dig, or tftp) to send UDP datagrams to target UDP network services and await a positive response.
Many security-conscious organizations filter ICMP messages to and from their Internet-based hosts, so it is often difficult to assess which UDP services are accessible via simple port scanning. If “ICMP destination port unreachable” messages can escape the target network, a traditional UDP port scan can be undertaken to identify open UDP ports on target hosts deductively.
UDP port scanning is an inverted scanning type in which open ports don’t respond. In particular, the scan looks for “ICMP destination port unreachable” (type 3 code 3) messages from the target host, as shown in Figure 4-13.
Nmap supports UDP port scanning with the
-sU option. SuperScan 4 also supports UDP
port scanning. However, both tools wait for negative “ICMP destination
port unreachable” messages to identify open ports (i.e., those ports
that don’t respond). If these ICMP messages are filtered by a firewall
as they try to travel out of the target network, the results will be
During a comprehensive audit of Internet-based network space,
you should send crafted UDP client packets to popular services and
await a positive response. The scanudp utility developed by Fryxar (http://www.geocities.com/fryxar) does this very well.
Example 4-7 shows scanudp being run against a Windows 2000
scanudpscanudp v2.0 - by: Fryxar usage: ./scanudp [options] <host> options: -t <timeout> Set port scanning timeout -b <bps> Set max bandwidth -v Verbose Supported protocol: echo daytime chargen dns tftp ntp ns-netbios snmp(ILMI) snmp(public) $
scanudp 192.168.0.50192.168.0.50 53 192.168.0.50 137 192.168.0.50 161
Use of fragmented probe packets that are assembled when they reach the target host
Use of spoofing to emulate multiple fake hosts launching network scanning probes, in which the real IP address of the scanning host is inserted to collect results
Filtering mechanisms can be circumvented at times using malformed or fragmented packets. However, the common techniques used to bypass packet filters at either the network or system kernel level are as follows:
Use of source routing
Use of specific TCP or UDP source ports
First, I’ll discuss IDS evasion techniques of fragmenting data and emulating multiple hosts, and then I’ll discuss filter circumvention methodologies. These techniques can often be mixed to launch attacks using source routed, fragmented packets to bypass both filters and IDS systems.
Probe packets can be fragmented easily with fragroute to fragment all probe packets flowing from your host or network or with a port scanner that supports simple fragmentation, such as Nmap. Many IDS sensors can’t process large volumes of fragmented packets because doing so creates a large overhead in terms of memory and CPU consumption at the network sensor level.
Dug Song’s fragtest utility (available as part of the fragroute package from http://www.monkey.org/~dugsong/fragroute) can determine exactly which types of fragmented ICMP messages are processed and responded to by the remote host. ICMP echo request messages are used by fragtest for simplicity and allow for easy analysis; the downside is that the tool can’t assess hosts that don’t respond to ICMP messages.
After undertaking ICMP probing exercises (such as ping sweeping and hands-on use of the sing utility) to ensure that ICMP messages are processed and responded to by the remote host, fragtest can perform three particularly useful tests:
Send an ICMP echo request message in 8-byte fragments
Send an ICMP echo request message in 8-byte fragments,
along with a 16-byte overlapping fragment, favoring newer data
in reassembly (using the
Send an ICMP echo request message in 8-byte fragments,
along with a 16-byte overlapping fragment, favoring older data
in reassembly (using the
Here is an example that uses fragtest to assess responses to
fragmented ICMP echo request messages with the
fragtest frag frag-new frag-old www.bbc.co.ukfrag: 467.695 ms frag-new: 516.327 ms frag-old: 471.260 ms
After ascertaining that fragmented and overlapped packets are indeed processed correctly by the target host and not dropped by firewalls or security mechanisms, a tool such as fragroute can be used to fragment all IP traffic destined for the target host.
The fragroute utility intercepts, modifies, and rewrites egress traffic destined for a specific host, according to a predefined rule set. When built and installed, version 1.2 comprises the following binary and configuration files:
The fragroute.conf file defines the way fragroute fragments, delays, drops, duplicates, segments, interleaves, and generally mangles outbound IP traffic.
Using the default configuration file, fragroute can be run from the command line in the following manner:
cat /usr/local/etc/fragroute.conftcp_seg 1 new ip_frag 24 ip_chaff dup order random print $
fragrouteUsage: fragroute [-f file] dst $
fragroute 192.168.102.251fragroute: tcp_seg -> ip_frag -> ip_chaff -> order -> print
Egress traffic processed by fragroute is displayed in tcpdump format if the
The fragroute man page
covers all the variables that can be set within the configuration
file. The type of IP fragmentation and reordering used by
fragtest when using the
frag-new option can be applied
to all outbound IP traffic destined for a specific host by
defining the following variables in the fragroute.conf file:
ip_frag 8 old order random print
TCP data can be segmented into 4-byte, forward-overlapping chunks (favoring newer data), interleaved with random chaff segments bearing older timestamp options (for PAWS elimination), and reordered randomly using these fragroute.conf variables:
tcp_seg 4 new tcp_chaff paws order random print
I recommend testing the variables used by fragroute in a controlled environment before live networks and systems are tested. This ensures that you see decent results when passing probes through fragroute and allows you to check for adverse reactions to fragmented traffic being processed. Applications and hardware appliances alike have been known to crash and hang from processing heavily fragmented and mangled data!
Nmap can fragment probe packets when launching half-open SYN or inverse TCP scanning types. The TCP header itself is split over several packets to make it more difficult for packet filters and IDS systems to detect the port scan. While most firewalls in high-security environments queue all the IP fragments before processing them, some networks disable this functionality because of the performance hit incurred. Example 4-8 shows Nmap being run to perform a fragment half-open SYN TCP scan.
nmap -sS -f 192.168.102.251Starting Nmap 4.10 ( http://www.insecure.org/nmap/ ) at 2007-04-01 23:25 UTC Interesting ports on cartman (192.168.102.251): (The 1524 ports scanned but not shown below are in state: closed) Port State Service 25/tcp open smtp 53/tcp open domain 8080/tcp open http-proxy
By emulating a large number of attacking hosts all launching probes and port scans against a target network, IDS alert and logging systems will effectively be rendered useless. Nmap allows for decoy hosts to be defined so that a target host can be scanned from a plethora of spoofed addresses (thus obscuring your own IP address).
The flag that defines decoy addresses within Nmap is
-D decoy1,ME,decoy2,decoyX. Example 4-9 shows Nmap being
used in this fashion to scan
nmap -sS -P0 -D 22.214.171.124,ME,126.96.36.199 192.168.102.251Starting Nmap 4.10 ( http://www.insecure.org/nmap/ ) at 2007-04-01 23:26 UTC Interesting ports on cartman (192.168.102.251): (The 1524 ports scanned but not shown below are in state: closed) Port State Service 25/tcp open smtp 53/tcp open domain 8080/tcp open http-proxy
Notice that the
-P0 flag is
also specified. When performing any kind of stealth attack it is
important that even initial probing (in the case of Nmap, an ICMP echo
request and attempted connection to TCP port 80) isn’t undertaken,
because it will reveal the true source of the attack in many
Source routing is a feature traditionally used for network troubleshooting purposes. Tools such as traceroute can be provided with details of gateways that the packet should be loosely or strictly routed through so that specific routing paths can be tested. Source routing allows you to specify which gateways and routes your packets should take, instead of allowing routers and gateways to query their own routing tables to determine the next hop.
Source routing information is provided as an IP options field in the packet header, as shown in Figure 4-14.
The format of the IP option data within a source-routed packet is quite simple. The first three bytes are reserved for IP option code, length, and pointer. Because IP option data can be used for different functionality (timestamp, strict routing, route, and record), the code field specifies the option type. The length field, oddly enough, states the size of the optional data, which can’t be larger than 40. Finally, the offset pointer field points to the current IP address in the remaining data section, which is rewritten as the packet traverses the Internet. Figure 4-15 shows the offset pointer in action.
There are two types of source routing, both defined in RFC 791:
Strict Source and Route Record (SSRR)
Loose Source and Route Record (LSRR)
Loose source routing allows the packet to use any number of intermediate gateways to reach the next address in the route. Strict source routing requires the next address in the source route to be on a directly connected network; if not, the delivery of the packet can’t be completed.
The source route options have a variable length, containing a series of IP addresses and an offset pointer indicating the next IP address to be processed. A source-routed datagram completes its delivery when the offset pointer points beyond the last field and the address in the destination address has been reached.
There is a limit of 40 characters for the router data within the IP options field. With 3 bytes used for the header information and 4 bytes committed for the final host address, there remain only 33 bytes to define loose hops, so 8 IP addresses can be defined in the list of hops (not counting the final destination host).
Source routing vulnerabilities can be exploited by:
Reversing the source route
Circumventing filters and gaining access to internal hosts
If a firewall or gateway reverses the source routing information when sending packets back, you can sniff traffic at one of the hops you defined. In a similar fashion to using sniffer-based spoofed scanning, you can launch scans and probes from potentially trusted hosts (e.g., branch office firewalls) and acquire accurate results.
In the case of Microsoft Windows NT hosts, the circumvention of filters involves manipulating the source routing options information to have an offset pointer set greater than the length of the list of hops and defining an internal host as the last hop (which is then reversed, sending the packet to the internal host). This vulnerability is listed in MITRE CVE (http://cve.mitre.org) as CVE-1999-0909.
A second source routing vulnerability (CVE-2006-2379) exists in the Windows TCP/IP driver for Windows 2003 SP1, Windows 2000 SP4, and Windows XP SP2 and earlier, which results in remote arbitrary code execution. Windows 2003 and XP are secure by default, as source routing support is disabled. At this time, however, there are no public exploit scripts available, although a simple DoS script can be found at http://www.milw0rm.com/exploits/1967.
Todd MacDermid of Syn Ack Labs (http://www.synacklabs.net) has written two excellent tools that can assess and exploit source routing vulnerabilities found in remote networks:
Both tools require libpcap and libdnet to build, and they run quite smoothly in Linux and BSD environments. A white paper written by Todd that explains source routing problems in some detail is available from http://www.synacklabs.net/oob/lsr.html. LSR attack mileage varies nowadays, as most ISPs drop LSR traffic, and so it does not usually traverse the Internet.
Whether the target host reverses the source route when sending packets back
Whether the target host can forward source-routed packets to an internal host, by setting the offset pointer to be greater than the number of hops defined in the loose hop list
The basic usage of the tool is as follows:
lsrscanusage: lsrscan [-p dstport] [-s srcport] [-S ip] [-t (to|through|both)] [-b host<:host ...>] [-a host<:host ...>] <hosts>
Some operating systems will reverse source-routed traffic only to ports that are open, so LSRScan should be run against an open port. By default, LSRScan uses a destination port of 80. The source port and source IP addresses aren’t necessary (LSRScan selects a random source port and IP address), but they can be useful in some cases.
-b option inserts IP
addresses of hops before the user’s host in the source route list,
-a option inserts
specific IP addresses after the user’s host in the list (although
those hosts must support source route forwarding for the scan to
be effective). For more information about the flags and options
that can be parsed, consult the LSRScan man page. Example 4-10 shows LSRScan being
run against a network block to identify hosts with source routing
lsrscan 188.8.131.52/24184.108.40.206 does not reverse LSR traffic to it 220.127.116.11 does not forward LSR traffic through it 18.104.22.168 reverses LSR traffic to it 22.214.171.124 forwards LSR traffic through it 126.96.36.199 reverses LSR traffic to it 188.8.131.52 does not forward LSR traffic through it
Because some systems reverse the source route, spoofing attacks using LSRTunnel can be performed. Knowing that systems forward source-routed traffic, accurate details of internal IP addresses have to be determined so that port scans can be launched through fragroute to internal space.
LSRTunnel spoofs connections using source-routed packets. For the tool to work, the target host must reverse the source route (otherwise the user will not see the responses and be able to spoof a full TCP connection). LSRTunnel requires a spare IP address on the local subnet to use as a proxy for the remote host.
Running LSRTunnel with no options shows the usage syntax:
lsrtunnelusage: lsrtunnel -i <proxy IP> -t <target IP> -f <spoofed IP>
The proxy IP is an unused network address an attacker uses to proxy connections between her host and the target address. The spoofed IP address is the host that appears as the originator of the connection. For additional details, consult the LSRTunnel manual page.
In this example of LSRTunnel,
192.168.102.2 is on the same local
subnet as the host:
lsrtunnel -i 192.168.102.2 -t 184.108.40.206 -f 220.127.116.11
At this point, LSRTunnel listens for traffic on the proxy IP
192.168.102.2). Using another
system on the network, any scan or attack traffic sent to the
proxy IP is forwarded to the target (
18.104.22.168) and rewritten to appear as
if it originated from relay2.ucia.gov (
When using a tool such as Nmap to perform either UDP or TCP port scanning of hosts, it is important to assess responses using specific source ports. Here are four source ports you should use along with UDP, half-open SYN, and inverse FIN scan types:
TCP or UDP port 53 (DNS)
TCP port 20 (FTP data)
TCP port 80 (HTTP)
TCP or UDP port 88 (Kerberos)
Using specific source ports, attackers can take advantage of firewall configuration issues and bypass filtering. UDP port 53 (DNS) is a good candidate when circumventing stateless packet filters because machines inside the network have to communicate with external DNS servers, which in turn respond using UDP port 53. Typically, a rule is put in place allowing traffic from UDP port 53 to destination port 53 or anything above 1024 on the internal client machine. Useful source ports to run scans from are TCP 20 (FTP data), UDP 53 (DNS), and UDP 500 (ISAKMP).
Check Point Firewall-1, Cisco PIX, and other stateful firewalls aren’t vulnerable to these issues (unless grossly misconfigured) because they maintain a state table and allow traffic back into the network only if a relative outbound connection or request has been initiated.
An inverse FIN scan should be attempted when scanning the HTTP service port because a Check Point Firewall-1 option known as fastmode is sometimes enabled for web traffic in high throughput environments (to limit use of firewall processing resources). For specific information regarding circumvention of Firewall-1 in certain configurations, consult the excellent presentation from Black Hat Briefings 2000 by Thomas Lopatic et al. titled “A Stateful Inspection of Firewall-1” (available as a Real Media video stream and PowerPoint presentation from http://www.blackhat.com/html/bh-usa-00/bh-usa-00-speakers.html).
On Windows 2000 and other Microsoft platforms that can run IPsec, a handful of default exemptions to the IPsec filter exist, including one that allows Kerberos (source TCP or UDP port 88) traffic into the host if the filter is enabled. These default exemptions are removed in Windows Server 2003, but they still pose a problem in some environments that rely on filtering at the operating system kernel level.
Tools such as Nmap, Hping2, and Firewalk perform low-level IP assessment. Sometimes holes exist to allow certain TCP services through the firewall, but the expected service isn’t running on the target host. Such low-level network details are useful to know, especially in sensitive environments (e.g., online banking environments), because very small holes in network integrity can sometimes be abused along with larger problems to gain or retain access to target hosts.
Insight into the following areas of a network can be gleaned through low-level IP assessment:
Uptime of target hosts (by analyzing the TCP timestamp option)
TCP services that are permitted through the firewall (by analyzing responses to TCP and ICMP probes)
TCP sequence and IP ID incrementation (by running predictability tests)
The operating system of the target host (using IP fingerprinting)
Nmap automatically attempts to calculate target host uptime information by analyzing the TCP timestamp option values of packets received. The TCP timestamp option is defined in RFC 1323; however, many platforms don’t adhere to RFC 1323. This feature often gives accurate results against Linux operating systems and others such as FreeBSD, but your mileage may vary.
A TCP probe always results in one of four responses. These responses potentially allow an analyst to identify where a connection was accepted, or why and where it was rejected, dropped, or lost:
If a SYN/ACK packet is received, the port is considered open.
If an RST/ACK packet is received, the probe packet was rejected by either the target host or an upstream security device (e.g., a firewall with a reject rule in its policy).
If an ICMP type 3 code 13 message is received, the host (or a device such as a firewall) has administratively prohibited the connection according to an Access Control List (ACL) rule.
If no packet is received, an intermediary security device silently dropped it.
Nmap returns details of ports that are open, closed, filtered, and unfiltered in line with this list. The unfiltered state is reported by Nmap from time to time, depending on the number of filtered ports found. If some ports don’t respond, but others respond with RST/ACK, the responsive ports are considered unfiltered (because the packet is allowed through the filter but the associated service isn’t running on the target host).
Hping2 can be used on a port-by-port basis to perform low-level analysis of responses to crafted TCP packets that are sent to destination network ports of remote hosts. Another useful tool is Firewalk, which performs filter analysis by sending UDP or TCP packets with specific TTL values. These unique features of Hping2 and Firewalk are discussed next.
Hping2 allows you to craft and send TCP packets to remote hosts with specific flags and options set. By analyzing responses at a low level, it is often possible to gain insight into the filter configuration at the network level. The tool is complex to use and has many possible options. Table 4-1 lists the most useful flags for performing low-level TCP assessment.
Send a specific number of probe packets
Source TCP port (random by default)
Destination TCP port
Set the TCP SYN flag
Set the TCP FIN flag
Set the TCP ACK flag
Here’s a best-practice use of Hping2 to assess a specific TCP port:
hping2 -c 3 -s 53 -p 139 -S 192.168.0.1HPING 192.168.0.1 (eth0 192.168.0.1): S set, 40 headers + 0 data ip=192.168.0.1 ttl=128 id=275 sport=139 flags=SAP seq=0 win=64240 ip=192.168.0.1 ttl=128 id=276 sport=139 flags=SAP seq=1 win=64240 ip=192.168.0.1 ttl=128 id=277 sport=139 flags=SAP seq=2 win=64240
In this example, a total of three TCP SYN packets are sent to
port 139 on
192.168.0.1 using the
source port 53 of the host (some firewalls ship with a configuration
that allows DNS traffic through the filter with an any-any rule, so
it is sometimes fruitful to use a source port of 53).
Following are four examples of Hping2 that generate responses in line with the four states discussed previously (open, closed, blocked, or dropped).
TCP port 80 is open:
hping2 -c 3 -s 53 -p 80 -S google.comHPING google.com (eth0 22.214.171.124): S set, 40 headers + 0 data ip=126.96.36.199 ttl=128 id=289 sport=80 flags=SAP seq=0 win=64240 ip=188.8.131.52 ttl=128 id=290 sport=80 flags=SAP seq=1 win=64240 ip=184.108.40.206 ttl=128 id=291 sport=80 flags=SAP seq=2 win=64240
TCP port 139 is closed or access to the port is rejected by a firewall:
hping2 -c 3 -s 53 -p 139 -S 192.168.0.1HPING 192.168.0.1 (eth0 192.168.0.1): S set, 40 headers + 0 data ip=192.168.0.1 ttl=128 id=283 sport=139 flags=R seq=0 win=64240 ip=192.168.0.1 ttl=128 id=284 sport=139 flags=R seq=1 win=64240 ip=192.168.0.1 ttl=128 id=285 sport=139 flags=R seq=2 win=64240
TCP port 23 is blocked by a router ACL:
hping2 -c 3 -s 53 -p 23 -S gw.example.orgHPING gw (eth0 192.168.0.254): S set, 40 headers + 0 data ICMP unreachable type 13 from 192.168.0.254 ICMP unreachable type 13 from 192.168.0.254 ICMP unreachable type 13 from 192.168.0.254
TCP probe packets are dropped in transit:
hping2 -c 3 -s 53 -p 80 -S 192.168.10.10HPING 192.168.10.10 (eth0 192.168.10.10): S set, 40 headers + 0 data
Mike Schiffman and Dave Goldsmith’s Firewalk utility (version 5.0 at the time of writing) allows assessment of firewalls and packet filters by sending IP packets with TTL values set to expire one hop past a given gateway. Three simple states allow you to determine if a packet has passed through the firewall or not:
If an ICMP type 11 code 0 (“TTL exceeded in transit”) message is received, the packet passed through the filter and a response was later generated.
If the packet is dropped without comment, it was probably done at the gateway.
If an ICMP type 3 code 13 (“Communication administratively prohibited”) message is received, a simple filter such as a router ACL is being used.
If the packet is dropped without comment, this doesn’t necessarily mean that traffic to the target host and port is filtered. Some firewalls know that the packet is due to expire and will send the “expired” message whether the policy allows the packet or not.
Firewalk works effectively against hosts in true IP routed environments, as opposed to hosts behind firewalls using network address translation (NAT). I recommend reading the Firewalk white paper written by Mike Schiffman and Dave Goldsmith, available from http://www.packetfactory.net/projects/firewalk/firewalk-final.pdf.
Example 4-11 shows Firewalk being run against a host to assess filters in place for a selection of TCP ports (21, 22, 23, 25, 53, and 80). The utility requires two IP addresses: the gateway (gw.test.org in this example) and the target (www.test.org in this example) that is behind the gateway.
firewalk -n -S21,22,23,25,53,80 -pTCP gw.test.org www.test.orgFirewalk 5.0 [gateway ACL scanner] Firewalk state initialization completed successfully. TCP-based scan. Ramping phase source port: 53, destination port: 33434 Hotfoot through 220.127.116.11 using 18.104.22.168 as a metric. Ramping Phase: 1 (TTL 1): expired [192.168.102.254] 2 (TTL 2): expired [22.214.171.124] 3 (TTL 3): expired [126.96.36.199] Binding host reached. Scan bound at 4 hops. Scanning Phase: port 21: A! open (port listen) [188.8.131.52] port 22: A! open (port not listen) [184.108.40.206] port 23: A! open (port listen) [220.127.116.11] port 25: A! open (port not listen) [18.104.22.168] port 53: A! open (port not listen) [22.214.171.124] port 80: A! open (port not listen) [126.96.36.199]
The tool first performs an effective traceroute to the target host in order to calculate the number of hops involved. Upon completing this initial reconnaissance, crafted TCP packets are sent with specific IP TTL values. By analyzing the responses from the target network and looking for ICMP type 11 code 0 messages, an attacker can reverse-engineer the filter policy of gw.test.org.
ICMP TTL exceeded (type 11 code 0) messages, indicating a routing loop
ICMP administratively prohibited (type 3 code 13) messages, indicating a firewall or router that rejects certain packets in line with an ACL
These ICMP response messages give insight into the target network’s setup and configuration. It is also possible to determine IP alias relationships in terms of firewalls performing NAT and other functions to forward traffic to other hosts and devices (for example, if you are probing a public Internet address but see responses from a private address in your sniffer logs).
Various operating platforms have their own interpretations of IP-related standards when receiving certain types of packets and responding to them. By carefully analyzing responses from Internet-based hosts, attackers can often guess the operating platform of the target host via IP fingerprinting, usually by assessing and sampling the following IP responses:
TCP FIN probes and bogus flag probes
TCP sequence number sampling
TCP WINDOW sampling
TCP ACK value sampling
ICMP message quoting
ICMP ECHO integrity
Responses to IP fragmentation
IP TOS (type of service) sampling
Originally, tools such as cheops and queso were developed specifically to guess target system operating platforms; however, the first publicly available tool to perform this was sirc3, which simply detected the difference between BSD-derived, Windows, and Linux TCP stacks.
Today, Nmap performs a large number of IP fingerprinting tests
to guess the remote operating platform. To enable IP fingerprinting
when running Nmap, simply use the
-O flag in combination with a scan type flag
-sS, as shown in Example 4-12.
nmap -O -sS 192.168.0.65Starting Nmap 4.10 ( http://www.insecure.org/nmap/ ) at 2007-04-01 23:26 UTC Interesting ports on 192.168.0.65: (The 1585 ports scanned but not shown below are in state: closed) Port State Service 22/tcp open ssh 25/tcp open smtp 53/tcp open domain 80/tcp open http 88/tcp open kerberos-sec 110/tcp open pop-3 135/tcp open loc-srv 139/tcp open netbios-ssn 143/tcp open imap2 389/tcp open ldap 445/tcp open microsoft-ds 464/tcp open kpasswd5 593/tcp open http-rpc-epmap 636/tcp open ldapssl 1026/tcp open LSA-or-nterm 1029/tcp open ms-lsa 1352/tcp open lotusnotes 3268/tcp open globalcatLDAP 3269/tcp open globalcatLDAPssl 3372/tcp open msdtc Remote OS guesses: Windows 2000 or WinXP
If TCP sequence numbers are generated in a predictable way by the target host, then blind spoofing and hijacking can occur (although this is usually limited to internal network spaces). Older Windows operating platforms suffer from this because the sequence numbers are simply incremented instead of randomly generated.
If the IP ID value is incremental, the host can be used as a third party to perform IP ID header scanning. IP ID header scanning requires the ID values returned from the third party to be incremental so that accurate scan results can be gathered.
Example 4-13 shows Nmap
being run in verbose mode (
TCP/IP fingerprinting (
both options shows the results of both TCP and IP ID sequence number
nmap -v -sS -O 192.168.102.251Starting Nmap 4.10 ( http://www.insecure.org/nmap/ ) at 2007-04-01 23:26 UTC Interesting ports on cartman (192.168.102.251): (The 1524 ports scanned but not shown below are in state: closed) Port State Service 25/tcp open smtp 53/tcp open domain 8080/tcp open http-proxy Remote OS guesses: Windows 2000 RC1 through final release TCP Sequence Prediction: Class=random positive increments Difficulty=15269 (Worthy challenge) IPID Sequence Generation: Incremental
Different IP network scanning methods allow you to test and effectively identify vulnerable network components. Here is a list of effective network scanning techniques and their applications:
By launching an ICMP ping sweep, you can effectively identify poorly protected hosts (as security-conscious administrators filter inbound ICMP messages) and perform a degree of operating system fingerprinting and reconnaissance by analyzing responses to the ICMP probes.
Inverse scanning types (particularly FIN, XMAS, and NULL) take advantage of idiosyncrasies in certain TCP/IP stack implementations. This scanning type isn’t effective when scanning large network spaces, although it is useful when testing and investigating the security of specific hosts and small network segments.
Using a combination of vulnerable network components and TCP spoofing, third-party TCP port scans can be effectively launched. Scanning in this fashion has two benefits: hiding the true source of a TCP scan and assessing the filters and levels of trust between hosts. Although time-consuming to undertake, third-party scanning is extremely useful when applied correctly.
Identifying accessible UDP services can be undertaken easily only if ICMP type 3 code 3 (“Destination port unreachable”) messages are allowed back through filtering mechanisms that protect target systems. UDP services can sometimes be used to gather useful data or directly compromise hosts (the DNS, SNMP, TFTP, and BOOTP services in particular).
Intrusion detection systems and other security mechanisms can be rendered ineffective by using multiple spoofed decoy hosts when scanning or by fragmenting probe packets using Nmap or fragroute. Filters such as firewalls, routers, and even software (including the Microsoft IPsec filter) can sometimes be bypassed using specific source TCP or UDP ports, source routing, or stateful attacks.
Here is a checklist of countermeasures to use when considering technical modifications to networks and filtering devices to reduce the effectiveness of network scanning and probing undertaken by attackers:
Filter inbound ICMP message types at border routers and firewalls. This forces attackers to use full-blown TCP port scans against all of your IP addresses to map your network correctly.
Filter all outbound ICMP type 3 “unreachable” messages at border routers and firewalls to prevent UDP port scanning and firewalking from being effective.
Consider configuring Internet firewalls so that they can identify port scans and throttle the connections accordingly. You can configure commercial firewall appliances (such as those from Check Point, NetScreen, and WatchGuard) to prevent fast port scans and SYN floods from being launched against your networks (however, this functionality can be abused by attackers using spoofed source addresses, resulting in DoS). On the open source side, there are many tools such as portsentry that can identify port scans and drop all packets from the source IP address for a given period of time.
Assess the way that your network firewall and IDS devices handle fragmented IP packets by using fragtest and fragroute when performing scanning and probing exercises. Some devices crash or fail under conditions in which high volumes of fragmented packets are being processed.
Ensure that your routing and filtering mechanisms (both firewalls and routers) can’t be bypassed using specific source ports or source routing techniques.
If you house publicly accessible FTP services, ensure that
your firewalls aren’t vulnerable to stateful circumvention attacks
relating to malformed
If a commercial firewall is in use, ensure the following:
The latest service pack is installed.
Antispoofing rules have been correctly defined so that the device doesn’t accept packets with private spoofed source addresses on its external interfaces.
Fastmode services aren’t used in Check Point Firewall-1 environments.
Investigate using reverse proxy servers in your environment if you require a high-level of security. A reverse proxy will not forward fragmented or malformed packets to the target systems, so a number of low-level attacks are thwarted.
Be aware of your own network configuration and its publicly accessible ports by launching TCP and UDP port scans along with ICMP probes against your own IP address space. It is surprising how many large companies still don’t properly undertake even simple port scanning exercises.