Firewall knock operator daemon
fwknopd is the server component for the FireWall Knock Operator, and is responsible for monitoring and processing Single Packet Authorization (SPA) packets that are generated by fwknop clients, modifying a firewall or ACL policy to allow the desired access after authenticating and decrypting a valid SPA packet (in that order), and removing access after a configurable timeout.
The main application of this program is to conceal services such as SSH with an additional layer of security in order to make the exploitation of vulnerabilities (both 0-day and unpatched code) much more difficult. In addition, services that are concealed in this fashion naturally cannot be scanned for with Nmap.
The main configuration for fwknopd is maintained within two files: fwknopd.conf and access.conf. The default location for these files is determined at package configuration (typically /etc/fwknop). The configuration variables within these files are described below.
Manually specify interface on which to sniff, e.g. “-i eth0”. This option is not usually needed because the “PCAP_INTF” keyword in the fwknopd.conf file defines the sniffing interface.
Run fwknopd in the foreground instead of becoming a daemon. When run in the foreground, message that would go to the log would instead be sent to stderr. This mode is usually used when testing and/or debugging.
List only firewall rules that any running fwknopd daemon has created and then exit.
Specify the location of the access.conf file. If this option is not given, fwknopd will use the compile-time default location (typically /etc/fwknop/access.conf).
Specify the location of the fwknopd.conf file. If this option is not given, fwknopd will use the default location (typically /etc/fwknop/fwknopd.conf.
Specify the number of candidate SPA packets to process and exit when this limit is reached.
Specify the location of the digest.cache file. If this option is not given, fwknopd will use the compile-time default location (typically /var/run/fwknop/digest.cache).
Dump the configuration values that fwknopd derives from the /etc/fwknop/fwknopd.conf' (or override files) and /etc/fwknop/access.conf on stderr.
List all firewall rules including those that have nothing to do with fwknopd.
Flush any firewall rules created by a running fwknopd process. This option allows the used to easily delete fwknopd firewall rules without having to wait for them to be timed out.
Kill the current fwknopd process. This provides a quick and easy way to stop fwknopd without having to look in the process table.
Set/override the system default locale setting.
Override config variable values that are normally read from the /etc/fwknop/fwknopd.conf' file with values from the specified file. Multiple override config files can be given as a comma-separated list.
Specify the location of the fwknopd.pid file. If this option is not given, fwknopd will use the compile-time default location (typically '/var/run/fwknop/fwknopd.pid).
Specify a Berkeley packet filter statement on the fwknopd command line. This overrides the value of the PCAP_FILTER variable taken from the /etc/fwknop/fwknopd.conf' file.
This option instructs fwknopd to read packet data from a pcap file instead of sniffing an interface directly. This mode is usually used for debugging purposes, and will disable SPA packet age checking unless it is manually enabled in the /etc/fwknop/fwknopd.conf file.
Allow fwknopd to sniff SPA packets regardless of whether they are received on the sniffing interface or sent from the sniffing interface. In the later case, this can be useful to have fwknopd sniff SPA packets that are forwarded through a system and destined for a different network. If the sniffing interface is the egress interface for such packets (and hence SPA packets are sent by this interface instead of received), then this option will need to used in order for fwknopd to see them. The default is to only sniff packets that are received on the sniffing interface. Note that this setting is independent of promiscuous mode.
Restart the currently running fwknopd processes. This option will preserve the command line options that were supplied to the original fwknopd process but will force fwknopd to re-read the fwknopd.conf and /etc/fwknop/access.conf files. This will also force a flush of the current “FWKNOP” iptables chain(s).
Rotate the digest cache file by renaming it to “<name>-old”, and starting a new one. The digest cache file is typically found in /var/run/fwknop/digest.cache.
Display the status of any fwknopd processes that may or not be running. If there is an existing fwknopd process then 0 is returned for the exit status and 1 is returned otherwise.
Allow messages to be sent to syslog even if the foreground mode is set.
Run fwknopd in verbose mode. This can option can be specified multiple times to increase the verbosity of the output to the system log file (or to the screen if running in the foreground).
Display usage information and exit.
Display version information and exit.
fwknopd references the /etc/fwknop/fwknopd.conf' file for configuration variables that define its operational parameters (what network interface and port to sniff, what features to enable/disable, etc.). The fwknopd.conf file does not define any access control directives.
The access control directives are contained in the /etc/fwknop/access.conf file. Access control directives define encryption keys and level of access that is granted to an fwknop client that has generated the appropriate encrypted SPA message.
This section list the more prominent configuration variables used by fwknopd. It is not a complete list. There are directives for the type of firewall used by fwknopd (i.e. iptables, ipfw, or pf). You will want to make sure to check these to make sure they have appropriate values. See the /etc/fwknop/fwknopd.conf' file for the full list and corresponding details.
Specify the ethernet interface on which fwknopd will sniff packets.
By default fwknopd puts the pcap interface into promiscuous mode. Set this to “N” to disable that behavior (non-promiscuous).
PCAP_FILTER <pcap filter spec>
Define the filter used for PCAP modes; fwknopd defaults to UDP port 62201. However, if an fwknop client uses the --rand-port option to send the SPA packet over a random port, then this variable should be updated to something like “udp dst portrange 10000-65535”.
This instructs fwknopd to not honor SPA packets that have an old time stamp. The value for “old” is defined by the “MAX_SPA_PACKET_AGE” variable. If “ENABLE_SPA_PACKET_AGING” is set to “N”, fwknopd will not use the client time stamp at all.
Defines the maximum age (in seconds) that an SPA packet will be accepted. This requires that the client system is in relatively close time synchronization with the fwknopd server system (NTP is good). The default age is 120 seconds (two minutes).
Defines an expiration date for the access stanza in MM/DD/YYYY format. All SPA packets that match an expired stanza will be ignored. This parameter is optional.
Defines an expiration date for the access stanza as the epoch time, and is useful if a more accurate expiration time needs to be given than the day resolution offered by the ACCESS_EXPIRE variable above. All SPA packets that match an expired stanza will be ignored. This parameter is optional.
Track digest sums associated with previous SPA packets processed by fwknopd. This allows digest sums to remain persistent across executions of fwknopd. The default is “Y”. If set to “N”, fwknopd will not check incoming SPA packet data against any previously save digests. It is a good idea to leave this feature on to reduce the possibility of being vulnerable to a replay attack.
Allow SPA clients to request access to services through an iptables firewall instead of just to it (i.e. access through the FWKNOP_FORWARD chain instead of the INPUT chain).
Allow SPA clients to request access to a local socket via NAT. This still puts an ACCEPT rule into the FWKNOP_INPUT chain, but a different port is translated via DNAT rules to the real one. So, the user would do “ssh -p <port>” to access the local service (see the --NAT-local and --NAT-rand-port on the fwknop client command line).
Set this to “Y” to enable a corresponding SNAT rule. By default, if forwarding access is enabled (see the “ENABLE_IPT_FORWARDING” variable above), then fwknopd creates DNAT rules for incoming connections, but does not also complement these rules with SNAT rules at the same time. In some situations, internal systems may not have a route back out for the source address of the incoming connection, so it is necessary to also apply SNAT rules so that the internal systems see the IP of the internal interface where fwknopd is running.
Specify the IP address for SNAT. This functionality is only enabled when “ENABLE_IPT_SNAT” is set to “Y” and by default SNAT rules are built with the MASQUERADE target (since then the internal IP does not have to be defined here in the /etc/fwknop/fwknopd.conf file), but if you want fwknopd to use the SNAT target, you must also define an IP address with the “SNAT_TRANSLATE_IP” variable.
Add ACCEPT rules to the FWKNOP_OUTPUT chain. This is usually only useful if there are no state tracking rules to allow connection responses out and the OUTPUT chain has a default-drop stance.
Specify the the maximum number of bytes to sniff per frame. 1500 is the default.
Flush all existing rules in the fwknop chains at fwknopd start time. The default is “Y”.
Flush all existing rules in the fwknop chains when fwknopd is stopped or otherwise exits cleanly. The default is “Y”.
If GPG keys are used instead of a Rijndael symmetric key, this is the default GPG keys directory. Note that each access block in /etc/fwknop/access.conf can specify its own GPG directory to override this default. If not set here or in an access.conf stanza, then the $HOME/.gnupg directory of the user running fwknopd (most likely root).
Set the locale (via the LC_ALL variable). This can be set to override the default system locale.
Allow fwknopd to acquire SPA data from HTTP requests (generated with the fwknop client in --HTTP mode). Note that when this is enabled, the “PCAP_FILTER” variable would need to be updated to sniff traffic over TCP/80 connections and a web server should be running on the same server as fwknopd.
Enable the fwknopd TCP server. This is a "dummy" TCP server that will accept TCP connection requests on the specified TCPSERV_PORT. If set to "Y", fwknopd will fork off a child process to listen for, and accept incoming TCP request. This server only accepts the request. It does not otherwise communicate. This is only to allow the incoming SPA over TCP packet which is detected via PCAP. The connection is closed after 1 second regardless. Note that fwknopd still only gets its data via pcap, so the filter defined by PCAP_FILTER needs to be updated to include this TCP port.
Sets the number of packets that are processed when the pcap_dispatch() call is made. The default is zero, since this allows fwknopd to process as many packets as possible in the corresponding callback where the SPA handling routine is called for packets that pass a set of prerequisite checks. However, if fwknopd is running on a platform with an old version of libpcap, it may be necessary to change this value to a positive non-zero integer. More information can be found in the pcap_dispatch(3) man page.
Sets the number of microseconds to passed as an argument to usleep() in the pcap loop. The default is 10000, or 1/10th of a second.
Controls whether fwknopd is permitted to sniff SPA packets regardless of whether they are received on the sniffing interface or sent from the sniffing interface. In the later case, this can be useful to have fwknopd sniff SPA packets that are forwarded through a system and destined for a different network. If the sniffing interface is the egress interface for such packets, then this variable will need to be set to "Y" in order for fwknopd to see them. The default is "N" so that fwknopd only looks for SPA packets that are received on the sniffing interface (note that this is independent of promiscuous mode).
Set the port number that the “dummy” TCP server listens on. This server is only spawned when “ENABLE_TCP_SERVER” is set to “Y”.
Override syslog identity on message logged by fwknopd. The defaults are usually ok.
Override syslog facility. The “SYSLOG_FACILITY” variable can be set to
This section describes the access control directives in the /etc/fwknop/access.conf file. Theses directives define encryption keys and level of access that is granted to fwknop clients that have generated the appropriate encrypted message.
The access.conf variables described below provide the access directives for the SPA packets with a source (or embedded request) IP that matches an address or network range defined by the “SOURCE” variable. All variables following “SOURCE” apply to the source stanza. Each “SOURCE” directive starts a new stanza.
This defines the source address from which the SPA packet will be accepted. The string “ANY” is also accepted if a valid SPA packet should be honored from any source IP. Every authorization stanza in /etc/fwknop/access.conf definition must start with the “SOURCE” keyword. Networks should be specified in CIDR notation (e.g. “192.168.10.0/24”), and individual IP addresses can be specified as well. Also, multiple IP\(cqs and/or networks can be defined as a comma separated list (e.g. “192.168.10.0/24,10.1.1.123”)
Define a set of ports and protocols (tcp or udp) that will be opened if a valid knock sequence is seen. If this entry is not set, fwknopd will attempt to honor any proto/port request specified in the SPA data (unless of it matches any “RESTRICT_PORTS” entries). Multiple entries are comma-separated.
Define a set of ports and protocols (tcp or udp) that are explicitly not allowed regardless of the validity of the incoming SPA packet. Multiple entries are comma-separated.
Define the symmetric key used for decrypting an incoming SPA packet that is encrypted by the fwknop client with Rijndael. The actual encryption key that is used is derived from the standard PBKDF1 algorithm. This variable is required for all SPA packets unless GnuPG is used instead (see the GPG variables below).
KEY_BASE64 <base64 encoded passphrase>
Same as the KEY option above, but specify the symmetric key as a base64 encoded string. This allows non-ascii characters to be included in the base64-decoded key.
Specify the HMAC key for authenticated encryption of SPA packets. This supports both Rijndael and GPG encryption modes, and is applied according to the encrypt-then-authenticate model.
HMAC_KEY_BASE64 <base64 encoded key>
Specify the HMAC key as a base64 encoded string. This allows non-ascii characters to be included in the base64-decoded key.
Define the length of time access will be granted by fwknopd through the firewall after a valid knock sequence from a source IP address. If “FW_ACCESS_TIMEOUT” is not set then the default timeout of 30 seconds will automatically be set.
Specify the encryption mode when AES is used. The default is CBC mode, but other modes can be selected such as OFB and CFB. In general, it is recommended to not use this variable and leave it as the default. Note that the string “legacy” can be specified in order to generate SPA packets with the old initialization vector strategy used by versions of fwknop before 2.5. With the 2.5 release, fwknop uses PBKDF1 for key derivation.
HMAC_DIGEST_TYPE <digest algorithm>
Specify the digest algorithm for incoming SPA packet authentication. Must be one of MD5, SHA1, SHA256, SHA384, or SHA512. This is an optional field, and if not specified then fwknopd defaults to using SHA256 if the access stanza requires an HMAC.
This instructs fwknopd to accept complete commands that are contained within an authorization packet. Any such command will be executed on the fwknopd server as the user specified by the “CMD_EXEC_USER” or as the user that started fwknopd if that is not set.
This specifies the user that will execute commands contained within a SPA packet. If not specified, fwknopd will execute it as the user it is running as (most likely root). Setting this to a non-root user is highly recommended.
Require a specific username from the client system as encoded in the SPA data. This variable is optional and if not specified, the username data in the SPA data is ignored.
Force all SPA packets to contain a real IP address within the encrypted data. This makes it impossible to use the -s command line argument on the fwknop client command line, so either -R has to be used to automatically resolve the external address (if the client behind a NAT) or the client must know the external IP and set it via the -a argument.
FORCE_NAT <IP> <PORT>
For any valid SPA packet, force the requested connection to be NAT\(cqd through to the specified (usually internal) IP and port value. This is useful if there are multiple internal systems running a service such as SSHD, and you want to give transparent access to only one internal system for each stanza in the access.conf file. This way, multiple external users can each directly access only one internal system per SPA key.
For any valid SPA packet, add an SNAT rule in addition to any DNAT rule created with a corresponding (required) FORCE_NAT variable. This is analogous to “SNAT_TRANSLATE_IP” from the /etc/fwknop/fwknopd.conf file except that it is per access stanza and overrides any value set with “SNAT_TRANSLATE_IP”. This is useful for situations where an incoming NAT\(cqd connection may be otherwise unanswerable due to routing constraints (i.e. the system receiving the SPA authenticated connection has a default route to a different device than the SPA system itself).
This is similar to the “FORCE_SNAT” variable, except that it is not necessary to also specify an IP address for SNAT rules because the MASQUERADE target is used instead.
Define the path to the GnuPG directory to be used by the fwknopd server. If this keyword is not specified within /etc/fwknop/access.conf then fwknopd will default to using the /root/.gnupg directory for the server key(s) for incoming SPA packets handled by the matching access.conf stanza.
Define a GnuPG key ID to use for decrypting SPA messages that have been encrypted by an fwknop client. This keyword is required for authentication that is based on GPG keys. The GPG key ring on the client must have imported and signed the fwknopd server key, and vice versa. It is ok to use a sensitive personal GPG key on the client, but each fwknopd server should have its own GPG key that is generated specifically for fwknop communications. The reason for this is that the decryption password for the server key must be placed within the /etc/fwknop/access.conf file for fwknopd to function (it has to be able to decrypt SPA messages that have been encrypted with the server\(cqs public key). For more information on using fwknop with GnuPG keys, see the following link: “http://www.cipherdyne.org/fwknop/docs/gpghowto.html”.
GPG_DECRYPT_PW <decrypt password>
Specify the decryption password for the gpg key defined by the “GPG_DECRYPT_ID” above. This is a required field for gpg-based authentication.
Allow fwknopd to leverage a GnuPG key pair that does not have an associated password. While this may sound like a controversial deployment mode, in automated environments it makes sense because "there is usually no way to store a password more securely than on the secret keyring itself" according to: “http://www.gnupg.org/faq/GnuPG-FAQ.html#how-can-i-use-gnupg-in-an-automated-environment”. Using this feature and removing the passphrase from a GnuPG key pair is useful in some environments where libgpgme is forced to use gpg-agent and/or pinentry to collect a passphrase.
With this setting set to Y, fwknopd check all GPG-encrypted SPA messages for a signature (signed by the sender\(cqs key). If the incoming message is not signed, the decryption process will fail. If not set, the default is N.
Setting this will allow fwknopd to accept incoming GPG-encrypted packets that are signed, but the signature did not pass verification (i.e. the signer key was expired, etc.). This setting only applies if the GPG_REQUIRE_SIG is also set to Y.
Define a list of gpg key ID\(cqs that are required to have signed any incoming SPA message that has been encrypted with the fwknopd server key. This ensures that the verification of the remote user is accomplished via a strong cryptographic mechanism. This setting only applies if the “GPG_REQUIRE_SIG” is set to Y. Separate multiple entries with a comma.
The main configuration file for fwknop.
Defines all knock sequences and access control directives.
fwknopd requires libfko which is normally included with both source and binary distributions, and is a dedicated library developed by the fwknop project.
For packet sniffing, fwknopd currently requires libpcap, but future versions still remove this as a dependency.
For GPG functionality, GnuPG must also be correctly installed and configured along with the libgpgme library.
To take advantage of all of the authentication and access management features of the fwknopd daemon/service a functioning iptables, ipfw, or pf firewall is required on the underlying operating system.
fwknopd can be run in debug mode by combining the -f, --foreground and the -v, --verbose command line options. This will disable daemon mode execution, and print verbose information to the screen on stderr as packets are received.
The most comprehensive way to gain diagnostic information on fwknopd is to run the test suite test-fwknop.pl script located in the test/ directory in the fwknop sources. The test suite runs sends fwknop through a large number of run time tests, has valgrind support, validates both SPA encryption and HMAC results against OpenSSL, and even has its own built in fuzzer for SPA communications.
More information on Single Packet Authorization can be found in the paper “Single Packet Authorization with fwknop” available at http://www.cipherdyne.org/fwknop/docs/SPA.html. A comprehensive tutorial on fwknop operations and theory can be found at http://www.cipherdyne.org/fwknop/docs/fwknop-tutorial.html. This tutorial also includes information about the design of fwknop that may be worth reading for those interested in why fwknop is different from other SPA implementations.
fwknop uses the git versioning system as its source code repository along with Github for tracking of issues and milestones:
$ git clone https://github.com/mrash/fwknop.git fwknop.git
Additional commentary on Single Packet Authorization can be found via Michael Rash\(cqs Twitter feed: http://twitter.com/michaelrash, @michaelrash
This “C” version of fwknop was derived from the original Perl-based version on which many people who are active in the open source community have contributed. See the CREDITS file in the fwknop sources, or visit http://www.cipherdyne.org/fwknop/docs/contributors.html to view the online list of contributors. A few contributors deserve to be singled out including: Franck Joncourt, Max Kastanas, Vlad Glagolev, Sean Greven, Hank Leininger, Fernando Arnaboldi, and Erik Gomez.
The phrase “Single Packet Authorization” was coined by MadHat and Simple Nomad at the BlackHat Briefings of 2005.
Send bug reports to [email protected] or [email protected], or open a new issue on Github (see https://github.com/mrash/fwknop.git). Suggestions and/or comments are always welcome as well. Additional information may be found in the fwknop mailing list archives (see: https://lists.sourceforge.net/lists/listinfo/fwknop-discuss).
fwknopd is distributed under the GNU General Public License (GPL), and the latest version may be downloaded from http://www.cipherdyne.org.