| TCP6(1) | General Commands Manual | TCP6(1) |
tcp6 - A security assessment tool for TCP/IPv6 implementations
tcp6 -i INTERFACE [-S LINK_SRC_ADDR] [-D LINK-DST-ADDR] [-s SRC_ADDR[/LEN]] [-d DST_ADDR] [-A HOP_LIMIT] [-y FRAG_SIZE] [-u DST_OPT_HDR_SIZE] [-U DST_OPT_U_HDR_SIZE] [-H HBH_OPT_HDR_SIZE] [-c OPEN_TYPE] [-C CLOSE_TYPE] [-P PAYLOAD_SIZE] [-o SRC_PORT] [-a DST_PORT] [-X TCP_FLAGS] [-q TCP_SEQ] [-Q TCP_ACK] [-V TCP_URP] [-w TCP_WIN] [-W WINDOW_MODE] [-M WIN_MOD_MODE] [-Z DATA] [-N] [-n] [-j PREFIX[/LEN]] [-k PREFIX[/LEN]] [-J LINK_ADDR] [-K LINK_ADDR] [-b PREFIX[/LEN]] [-g PREFIX[/LEN]] [-B LINK_ADDR] [-G LINK_ADDR] [-F N_SOURCES] [-T N_PORTS] [-f] [-R] [-L] [-l] [-p PROBE_MODE] [-z SECONDS] [-r RATE] [-v] [-h]
tcp6 tool allows the assessment of IPv6 implementations with respect to a variety of attack vectors based on TCP/IPv6 segments. This tool is part of the SI6 Networks' IPv6 Toolkit: a security assessment suite for the IPv6 protocols.
tcp6 can be employed to perform specific TCP connection-establishment and connection-termination sequences. For example, it can be employed to simulate a TCP "simultaneous open" scenario (see the "--open-mode" option) or to simulate a TCP "simultaneous close" scenario (see the "--close-mode" option). It can be employed to perform a number of resource-exhaustion attacks against TCP, such as SYN-flooding attacks, connection-flooding attacks (see the "--flood-sources" and "--flood-ports" options), etc. Additionally, it can be employed to exploit attack vectors based on the TCP window (see the "--window" option).
tcp6 tool has two modes of operation: active and listening. In active mode, the tool attacks a specific target, while in listening mode the tool listens to TCP traffic on the local network, and launches an attack in response to such traffic. Active mode is employed if an IPv6 Destination Address is specified. Listening mode is employed if the "-L" option (or its long counterpart "--listen") is set. If both an attack target and the "-L" option are specified, the attack is launched against the specified target, and then the tool enters listening mode to respond incoming packets with TCP segments.
tcp6 supports filtering of incoming packets based on the Ethernet Source Address, the Ethernet Destination Address, the IPv6 Source Address, and the IPv6 Destination Address. There are two types of filters: "block filters" and "accept filters". If any "block filter" is specified, and the incoming packet matches any of those filters, the message is discarded (and thus no TCP segments are sent in response). If any "accept filter" is specified, incoming packets must match the specified filters in order for the tool to respond with TCP segments.
tcp6 takes itS parameters as command-line options. Each of the options can be specified with a short name (one character preceded with the hyphen character, as e.g. "-i") or with a long name (a string preceded with two hyphen characters, as e.g. "--interface").
If the tool is instructed to e.g. flood the victim with TCP segments from different sources ("--flood-sources" option), multiple packets may need to be generated.
tcp6 supports IPv6 Extension Headers, including the IPv6 Fragmentation Header, which might be of use to circumvent layer-2 filtering and/or Network Intrusion Detection Systems (NIDS). However, IPv6 extension headers are not employed by default, and must be explicitly enabled with the corresponding options.
This option specifies the link-layer Source Address of the probe packets. If left unspecified, the link-layer Source Address of the packets is set to the real link-layer address of the network interface. Note: this option is meaningful only when the underlying link-layer technology is Ethernet.
This option specifies the link-layer Destination Address of the probe packets. By default, the link-layer Destination Address is automatically set to the link-layer address of the destination host (for on-link destinations) or to the link-layer address of the first-hop router. Note: this option is meaningful only when the underlying link-layer technology is Ethernet.
This option specifies the IPv6 source address (or IPv6 prefix) to be used for the Source Address of the attack packets. If the "-F" ("--flood-sources") option is specified, this option includes an IPv6 prefix, from which random addresses are selected. See the description of the "-F" option for further information on how the "-s" option is processed in that specific case.
Note: When operating in "listening" mode, the Source Address is automatically set to the Destination Address of the incoming packet.
This option specifies the IPv6 Destination Address of the victim. It can be left unspecified only if the "-L" option is selected (i.e., if the tool is to operate in "listening" mode).
Note: When operating in "listening" mode, the Destination Address is automatically set to the Source Address of the incoming packet.
This option specifies the Hop Limit to be used for the IPv6 packets. It defaults to 255.
This option specifies that a Destination Options header is to be included in the outgoing packet(s). The extension header size must be specified as an argument to this option (the header is filled with padding options). Multiple Destination Options headers may be specified by means of multiple "-u" options.
This option specifies a Destination Options header to be included in the "unfragmentable part" of the outgoing packet(s). The header size must be specified as an argument to this option (the header is filled with padding options). Multiple Destination Options headers may be specified by means of multiple "-U" options.
This option specifies that a Hop-by-Hop Options header is to be included in the outgoing packet(s). The header size must be specified as an argument to this option (the header is filled with padding options). Multiple Hop-by-Hop Options headers may be specified by means of multiple "-H" options.
This option specifies that the resulting packet must be fragmented. The fragment size must be specified as an argument to this option.
This options specifies the size of the TCP payload. It defaults to 0 (i.e., empty TCP segments).
This option specifies the TCP Source Port.
This option specifies the TCP/UDP Destination Port.
his option is used to set specific the TCP flags. The flags are specified as "F" (FIN), "S" (SYN), "R" (RST), "P" (PSH), "A" (ACK), "U" (URG), "X" (no flags).
If this option is not set, and the tool operates in listening mode, the flags of the generated TCP segments are automatically set as follows: TCP segments elicited by SYNs have both the SYN and ACK flags set. All other TCP segments have the ACK bit set.
This option specifies the Sequence Number of the TCP header. If left unspecified, the Sequence Number is randomized.
If this option is left unspecified and the tool is operating in listening mode, the TCP Sequence Number is set to the Acknowledgement Number of the packet that elicited the TCP segment.
This option specifies the Acknowledgment Number of the TCP segment. If left unspecified, the Acknowledgment Number is randomized.
If this option is left unspecified and the tool is operating in listening mode, the TCP Sequence Number is set to the Acknowledgement Number of the packet that elicited the TCP segment.
This option specifies the Urgent Pointer of the TCP segment. If left unspecified, the Urgent Pointer is set to 0.
This option specifies the value of the TCP Window. If left unspecified, the Window is randomized.
This option specifies how to operate the TCP window by means of the WIN_MODE parameter. Two modes are supported:
+ closed
+ modulated
When the "closed" mode is selected, the TCP window will be set to 0 (i.e., "closed window"). If the tool estabishes new TCP connections, the initial window advertised during the TCP three-way handshake will be that specified with the '-w' option. However, once the connection has been established, the TCP window will be set to 0. This allows for the implementation of the so-called Netkill attack, discussed in Section 7.1.1 of the document "Security Assessment of the Transmission Control Protocol (TCP)" (available at: <http://www.gont.com.ar/papers/tn-03-09-security-assessment-TCP.pdf>).
When the "modulated" mode is selected, the TCP window will oscillate between alternate between two different values. These values, along the amount of time that each of them is "active", can be specified by means of the '-M' ("--win-modulate") option. The first of the aforementioned values is meant to close the window (hence it will typically be zero), while the second is meant to open the window. The goal of alternating between these two values is to circumvent a trivial mitigation against Zero-Window attacks implemented by some stacks where they enforce a limit on the maximum amount of time that the TCP advertised by a remote peer remains fully-closed (i.e., set to 0). By changing the advertised window to some other (small) value every now and then, such a trivial "counter-measure" can be easily circumvented.
This option specifies the two values (and their respective duration) over which the TCP window will alternate. The value WIN_MOD_MODE hast the syntax "WIN1:TIME1:WIN2:TIME2", where the WIN1 and WIN2 parameters specify the window size for each of these periods, while the TIME1 and TIME2 parameters specify their respective time lengths. For example, setting "--win-modulation 0:60:10:30" will cause tcp6 to alternate between advertising a TCP window of 0 bytes for 60 seconds, and advertising a TCP window of 10 bytes for 30 seconds.
This option will be typically employed along with one of the flooding options ("--flood-sources" an/or "--flood-ports") and the "--data" option, such that multiple TCP connections are established, and the target TCPs keep their retransmission buffer full. In this scenario, the TCP window "modulation" option can be leveraged to evade trivial counter-measures implemented by some TCP stacks that try to mitigate Zero-Window attacks by enforcing a limit on the maximum amount of time the TCP window can be in the "closed state".
This option specifies the connection-establishment mode. The following modes are available:
+ simultaneous
+ passive
+ abort
When the "simultaneous" mode is selected, tcp6 will respond to incoming SYN segments with other SYN segments, thus simulating a "simultaneous open" scenario. When the "passive" mode is selected, tcp6 will respond to incoming SYN segments with the typical SYN/ACK segments, thus leading to the traditional "three-way handshake". Finally, when the "abort" mode si selected, tcp6 wil respond to incoming SYN segments with RST segments, thus aborting the incoming connections.
For the most part, this option is useful for assessing the correct behavior of TCP implementations (e.g., support for "simultaneous opens").
This option specifies the the connection-termination mode. The following modes are available:
+ simultaneous
+ passive
+ abort
+ active
+ FIN-WAIT-1
+ FIN-WAIT-2
+ LAST-ACK
When the "simultaneous" mode is selected, tcp6 will respond to incoming FIN segments with FIN segments, thus simulating a "simultaneous close" scenario. When the "passive" mode is selected, tcp6 will respond to incoming FIN segments with the typical FIN/ACK segments, thus leading to the traditional TCP connection-termination sequence. When the "abort" mode is selected, tcp6 wil respond to incoming FIN segments with RST segments, thus aborting the corresponding connections. When the "active" mode is selected, tcp6 will start the connection-termination sequence by sending a FIN segment.
The FIN-WAIT-1, FIN-WAIT-2, and LAST-ACK modes will result in connections in the FIN-WAIT-1, FIN-WAIT-2, and LAST-ACK, respectively. It should be noted that in order for the remote TCPs to transition to the FIN-WAIT-1 or FIN-WAIT-2 states, the remote TCPs must perform the "active close". This can be trivially triggered for application protocols such as HTTP, but might not be feasible for other protocols.
This option is used to specify a payload that should be sent as the first data segment once a TCP connection has been established. It will typically include an application-layer request. Note: the string used for the DATA parameter can contain the "\r" and "\n" C-style escape senquenced for representing "carriage return" and "line feed" (respectively).
As an example, this option could be employed to send an HTTP request if set as '--data "GET / HTTP/1.0\r\n\r\n"'.
This option instructs tcp6 not to acknowledge the TCP payload of incoming segments (when operating in listening mode).
Note: By default, tcp6 will acknowledge both the payload and the flags of the incoming TCP segments.
This option instructs tcp6 not to acknowledge the TCP flags (SYN and/or FIN) of incoming segments (when operating in listening mode).
Note: By default, tcp6 will acknowledge both the payload and the flags of the incoming TCP segments.
This option sets a block filter for the incoming packets, based on their IPv6 Source Address. It allows the specification of an IPv6 prefix in the form "-j prefix/prefixlen". If the prefix length is not specified, a prefix length of "/128" is selected (i.e., the option assumes that a single IPv6 address, rather than an IPv6 prefix, has been specified).
This option sets a block filter for the incoming packets, based on their IPv6 Destination Address. It allows the specification of an IPv6 prefix in the form "-k prefix/prefixlen". If the prefix length is not specified, a prefix length of "/128" is selected (i.e., the option assumes that a single IPv6 address, rather than an IPv6 prefix, has been specified).
This option sets a block filter for the incoming packets, based on their link-layer Source Address. The option must be followed by a link-layer address (currently, only Ethernet is supported).
This option sets a block filter for the incoming packets, based on their link-layer Destination Address. The option must be followed by a link-layer address (currently, only Ethernet is supported).
This option sets an accept filter for the incoming packets, based on their IPv6 Source Address. It allows the specification of an IPv6 prefix in the form "-b prefix/prefixlen". If the prefix length is not specified, a prefix length of "/128" is selected (i.e., the option assumes that a single IPv6 address, rather than an IPv6 prefix, has been specified).
This option sets a accept filter for the incoming packets, based on their IPv6 Destination Address. It allows the specification of an IPv6 prefix in the form "-g prefix/prefixlen". If the prefix length is not specified, a prefix length of "/128" is selected (i.e., the option assumes that a single IPv6 address, rather than an IPv6 prefix, has been specified).
This option sets an accept filter for the incoming packets, based on their link-layer Source Address. The option must be followed by a link-layer address (currently, only Ethernet is supported).
This option sets an accept filter for the incoming packets, based on their link-layer Destination Address. The option must be followed by a link-layer address (currently, only Ethernet is supported).
This option instructs the tool to send multiple TCP segments with different Source Addresses. The number of different source addresses is specified as "-F number". The Source Address of each TCP segment is randomly selected from the prefix specified by the "-s" option. If the "-F" option is specified but the "-s" option is left unspecified, the Source Address of the packets is randomly selected from the prefix ::/0.
This option instructs the tool to send multiple TCP segments with different Source Ports. The Source Port of each TCP segment is randomly selected from the whole port number space (0-65535).
This option instructs the tcp6 tool to send periodic TCP segments to the victim node. The amount of time to pause between sending TCP segments can be specified by means of the "-z" option, and defaults to 1 second. Note that this option cannot be set in conjunction with the "-L" ("--listen") option.
This option specifies the amount of time to pause between sending TCP segments (when the "--loop" option is set). If left unspecified, it defaults to 1 second.
This option specifies the rate limit to use when performing a remote address scan. "RATE" should be specified as "xbps" or "xpps" (with "x" being an unsigned integer), for rate-limits in bits per second or packets per second, respectively.
This instructs the tcp6 tool to operate in listening mode (possibly after attacking a given node). Note that this option cannot be used in conjunction with the "-l" ("--loop") option.
This option instructs tcp6 to operate in probe mode. The specific probe mode is specified as an argument to this option (currently, only "script" mode is supported). In probe mode, tcp6 sends probe segments, and waits for response packets. The response packets are decoded based on the selected probe mode.
In the "script" probe mode, the tool decodes TCP segments as follows:
RESPONSE:RESPONSE_TYPE:RESPONSE_DECODE...
Where the string RESPONSE is fixed, and RESPONSE_TYPE indicates the response received. As of this version of the tool, the following RESPONSE_TYPE values are supported:
+ TCP6: Indicates that the tool received a TCP/IPv6 packet
+ TIMEOUT: Indicates that the tool received no response
If RESPONSE_TYPE is TCP6, RESPONSE code contains the TCP flags set in the receive TCP segment. The TCP flags are encoded as "F" (FIN), "S" (SYN), "R" (RST), "P" (PSH), "A" (ACK), and "U" (URG).
Possibe output lines of the tool are:
RESPONSE:TIMEOUT:
RESPONSE:TCP6:RA:
Note: Future versions of the tool will also decode ICMPv6 error messages, and will include additional data regarding the incoming TCP segments (e.g., ACK value, payload size, etc.).
This option instructs the tcp6 tool to be verbose. When the option is set twice, the tool is "very verbose", and the tool also informs which packets have been accepted or discarded as a result of applying the specified filters.
Print help information for the tcp6 tool.
The following sections illustrate typical use cases of the tcp6 tool.
Example #1
# tcp6 -s fc00:1::/64 -d fc00:1::1 -a 22 -X S -F 100 -l -z 1 -v
In this example the tcp6 tool is essentially employed to perform a SYN-flood attack against port number 22 of the host fc00:1::1. The tool uses the network interface "eth0" (as specified by the "-i" option), and sends SYN segments (as specified by the "-X" option) from the prefix fc00:1::/64 (as specified by the "-s" option) to port 22 (specified by the "-a" option) at the destination address fc00:1::1 (specified by the "-d" option). The tool sends TCP segments from 100 different addresses (as specified by the "-F" option) every one second (as specified by the "-l" and "-z" options). The tool will be verbose (as specified by the "-v" option).
Example #2
# tcp6 -i eth0 -L -X RA -v
In this example, the tcp6 tool is employed to perform a TCP connection-reset attack against all active TCP connections in the local network. The tool listens ("-L") on the interface eth0 ("-i eth0"), and responds to any TCP segments with a RST packet (with both the RST and ACK bits set). The tool will be verbose.
Example #3
# tcp6 -i eth0 -d fc00:1::1 -a 80 -L -s fc00:1::/112 -l -r 1pps -v --data "GET / HTTP/1.0\r\n\r\n" --close-mode last-ack --flood-ports 10
Flood the target system (fc00:1::1) with connections that stay in the LAST-ACK state (on port 80), sending packets at a rate of one packet per second. For each forged address, 10 different (forged) ports are used. For each connection, tcp6 will send an HTTP application request.
Example #4
# tcp6 -i eth0 -d fc00:1::1 -a 80 -L -s fc00:1::/112 -l -r 1000pps --tcp-flags auto -v --data "GET / HTTP/1.0\r\n\r\n" --flood-ports 10 --window-mode close
Flood the target node (fc00:1::1) with TCP connections (on port 80). On each connection that is established, an HTTP request is sent, and the TCP window is immediately closed. For each forged IPv6 source address ten different TCP source ports are randomized. The bandwidth of the attack is limited to 1000 pps.
Example #5
# tcp6 -d fc00:1::1 -a 80 --tcp-flags A --dst-opt-hdr 8 --payload-size 50 --probe-mode script
Send a probe TCP segment to TCP port 80 at fc00:1::1. The probe packet consists of an IPv6 packet with a Destination Options header of 8 bytes, and an IPv6 payload consisting of a TCP segment with the ACK bit set, and 50 data bytes. The probe mode is "script".
"Security Assessment of the Transmission Control Protocol (TCP)" (available at: <http://www.gont.com.ar/papers/tn-03-09-security-assessment-TCP.pdf>) for a discussion of TCP vulnerabilities.
The tcp6 tool and the corresponding manual pages were produced by Fernando Gont <fgont@si6networks.com> for SI6 Networks <http://www.si6networks.com>.
Copyright (c) 2011-2013 Fernando Gont.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is available at <http://www.gnu.org/licenses/fdl.html>.