Writing Snort Rules
How To write Snort rules and keep your sanity
Current as of version 1.6
by Martin Roesch
Contents

The Basics
Snort uses a simple, lightweight rules description language that is flexible and quite powerful.  There are a number of simple guidelines to remember when developing Snort rules.

The first is that Snort rules must be completely contained on a single line, the Snort rule parser doesn't know how to handle rules on multiple lines.

Snort rules are divided into two logical sections, the rule header and the rule options.  The rule header contains the rule's action, protocol, source and destination IP addresses and netmasks, and the source and destination ports information.  The rule option section contains alert messages and information on which parts of the packet should be inspected to determine if the rule action should be taken.

Here is an example rule:

alert tcp any any -> 192.168.1.0/24 111 (content:"|00 01 86 a5|"; msg: "mountd access";)
Figure 1 - Sample Snort Rule

The text up to the first parenthesis is the rule header and the section enclosed in parenthesis is the rule options.  The words before the colons in the rule options section are called option keywords.  Note that the rule options section is not specifically required by any rule, they are just used for the sake of making tighter definitions of packets to collect or alert on (or drop, for that matter).  All of the elements in that make up a rule must be true for the indicated rule action to be taken.  When taken together, the elements can be considered to form a logical AND statement.  At the same time, the various rules in a Snort rules library file can be considered to form a large logical OR statement.  Let's begin by talking about the rule header section.
Rule Headers
Rule Actions:

The rule header contains the information that defines the "who, where, and what" of a packet, as well as what to do in the event that a packet with all the attributes indicated in the rule should show up.  The first item in a rule is the rule action.  The rule action tells Snort what to do when it finds a packet that matches the rule criteria.  There are three available actions in Snort, alert, log, and pass.

Protocols:

The next field in a rule is the protocol.  There are three IP protocols that Snort currently analyzes for suspicious behavior, tcp, udp, and icmp.  In the future there may be more, such as ARP, IGRP, GRE, OSPF, RIP, IPX, etc.

IP Addresses:

The next portion of the rule header deals with the IP address and port information for a given rule.  The keyword "any" may be used to define any address.  Snort does not have a mechanism to provide host name lookup for the IP address fields in the rules file.  The addresses are formed by a straight numeric IP address and a CIDR block.  The CIDR block indicates the netmask that should be applied to the rule's address and any incoming packets that are tested against the rule.  A CIDR block mask of /24 indicates a Class C network, /16 a Class B network, and /32 indicates a specific machine address.  For example, the address/CIDR combination 192.168.1.0/24 would signify the block of addresses from 192.168.1.1 to 192.168.1.255.  Any rule that used this designation for, say, the destination address would match on any address in that range.  The CIDR designations give us a nice short-hand way to designate large address spaces with just a few characters.

In Figure 1, the source IP address was set to match for any computer talking, and the destination address was set to match on the 192.168.1.0 Class C network.

There is an operator that can be applied to IP addresses, the negation operator.  This operator tells Snort to match any IP address except the one indicated by the listed IP address.  The negation operator is indicated with a "!".  For example, an easy modification to the initial example is to make it alert on any traffic that originates outside of the local net with the negation operator as shown in Figure 2.

alert tcp !192.168.1.0/24 any -> 192.168.1.0/24 111 (content: "|00 01 86 a5|"; msg: "external mountd access";)
Figure 2 - Example IP Address Negation Rule

This rule's IP addresses indicate "any tcp packet with a source IP address not originating from the internal network and a destination address on the  internal network".

Port Numbers

Port numbers may be specified in a number of ways, including "any" ports, static port definitions, ranges, and by negation.  "Any" ports are a wildcard value, meaning literally any port.  Static ports are indicated by a single port number, such as 111 for portmapper, 23 for telnet, or 80 for http, etc.  Port ranges are indicated with the range operator ":".  The range operator may be applied in a number of ways to take on different meanings, such as in Figure 3.
 
 

log udp any any -> 192.168.1.0/24 1:1024 
log udp traffic coming from any port and destination ports ranging from 1 to 1024
log tcp any any -> 192.168.1.0/24 :6000 
log tcp traffic from any port going to ports less than or equal to 6000
log tcp any :1024 -> 192.168.1.0/24 500: 
log tcp traffic from priveleged ports less than or equal to 1024 going to ports greater than or equal to 500
Figure 3 - Port Range Examples


Port negation is indicated by using the negation operator "!".  The negation operator may be applied against any of the other rule types (except any, which would translate to none, how Zen...).  For example, if for some twisted reason you wanted to log everything except the X Windows ports, you could do something like the rule in Figure 4.
 
 

log tcp any any -> 192.168.1.0/24 !6000:6010
Figure 4 - Example of Port Negation


The Direction Operator

The direction operator "->" indicates the orientation, or "direction", of the traffic that the rule applies to.  The IP address and port numbers on the left side of the direction operator is considered to be the traffic coming from the source host, and the address and port information on the right side of the operator is the destination host.  There is also a bidirectional operator, which is indicated with a "<>" symbol.  This tells Snort to consider the address/port pairs in either the source or destination orientation.  This is handy for recording/analyzing both sides of a conversation, such as telnet or POP3 sessions.  An example of the bidirectional operator being used to record both sides of a telnet session is shown in Figure 5.
 
 

log !192.168.1.0/24 any <> 192.168.1.0/24 23
Figure 5 - Snort rules using the Bidirectional Operator
Rule Options
Rule options form the heart of Snort's intrusion detection engine, combining ease of use with power and flexibility.  All Snort rule options are separated from each other using the semicolon ";" character.  Rule option keywords are separated from their arguments with a colon ":" character. As of this writing, there are fifteen rule option keywords  available for Snort:
Msg

The msg rule option tells the logging and alerting engine the message to print along with a packet dump or to an alert.  It is a simple text string that utilizes the "\" as an escape character to indicate a discrete character that might otherwise confuse Snort's rules parser (such as the semi-colon ";" character).

Format:

msg: "<message text>";
Logto

The logto option tells Snort to log all packets that trigger this rule to a special output log file.  This is especially handy for combining data from things like NMAP activity, HTTP CGI scans, etc.  It should be noted that this option does not work when Snort is in binary logging mode.

Format:

logto: "<filename>";
TTL

This rule option is used to set a specific time-to-live value to test against.  The test it performs is only sucessful on an exact match.  This option keyword was intended for use in the detection of traceroute attempts.

Format:

ttl: "<number>";
ID

This option keyword is used to test for an exact match in the IP header fragment ID field.  Some hacking tools (and other programs) set this field specifically for various purposes, for example the value 31337 is very popular with some hackers.  This can be turned against them by putting a simple rule in place to test for this and some other "hacker numbers".

Format:

id: "<number>";
Dsize

The dsize option is used to test the packet payload size.  It may be set to any value, plus use the greater than/less than signs to indicate ranges and limits.  For example, if you know that a certain service has a buffer of a certain size, you can set this option to watch for attempted buffer overflows.  It has the added advantage of being a much faster way to test for a buffer overflow than a payload content check.

Format:

dsize: [>|<] <number>;
Note: The > and < operators are optional!
Content

The content keyword is one of the more important features of Snort.  It allows the user to set rules that search for specific content in the packet payload and trigger response based on that data.  Whenever a content option pattern match is performed, the Boyer-Moore pattern match function is called and the (rather computationally expensive) test is performed against the packet contents.  If data exactly matching the argument data string os contained anywhere within the packet's payload, the test is successful and the remainder of the rule option tests are performed.  Be aware that this test is case sensitive.

The option data for the content keyword is somewhat complex; it can contain mixed text and binary data.  The binary data is generally enclosed within the pipe ("|") character and represented as bytecode.  Bytecode represents binary data as hexidecimal numbers and is a good shorthand method for describing complex binary data.  Figure 7 contains an example of mixed text and binary data in a Snort rule.
 
 

alert tcp any any -> 192.168.1.0/24 143 (content: "|90C8 C0FF FFFF|/bin/sh"; msg: "IMAP buffer overflow!";)
Figure 7 - Mixed Binary Bytecode and Text in a Content Rule Option

Format:

content: "<content string>";
Offset

The offset rule option is used as a modifier to rules using the content option keyword.  This keyword modifies the starting search position for the pattern match function from the beginning of the packet payload.  It is very useful for things like CGI scan detection rules where the content search string is never found in the first four bytes of the payload.  Care should be taken against setting the offset value too "tightly" and potentially missing an attack!  This rule option keyword cannot be used without also specifying a content rule option.

Format:

offset: <number>;
Depth

Depth is another content rule option modifier.  This sets the maximum search depth for the content pattern match function to search from the beginning of its search region.  It is useful for limiting the pattern match function from performing inefficient searches once the possible search region for a given set of content has been exceeded.  (Which is to say, if you're searching for "cgi-bin/phf" in a web-bound packet, you probably don't need to waste time searching the payload beyond the first 20 bytes!)  See Figure 8 for an example of a combined content, offset, and depth search rule.

Format:

depth: <number>;

 
alert tcp any any -> 192.168.1.0/24 80 (content: "cgi-bin/phf"; offset: 3; depth: 22; msg: "CGI-PHF access";)
Figure 8 - Combined Content, Offset and Depth Rule
Nocase

The nocase option is used to deactivate case sensitivity in a "content" rule.  It is specified alone within a rule and any ASCII characters that are compared to the packet payload are treated as though they are either upper of lower case.

Format:

nocase;

 
alert tcp any any -> 192.168.1.0/24 21 (content: "USER root"; nocase; msg: "FTP root user access attempt";)
Figure 9 - Content rule with nocase modifier
Flags

This rule tests the TCP flags for an exact match.  There are actually 8 flags variables available in Snort:

The reserved bits can be used to detect unusual behavior, such as IP stack fingerprinting attempts or other suspicious activity.  All of the flags are considered as a whole for this test, they must all be "up" for this rule option to be successful.  For instance, Figure 9 shows a SYN-FIN scan detection rule.

Format:

flags: <flag values>;

 
alert any any -> 192.168.1.0/24 any (flags: SF; msg: "Possible SYN FIN scan";)
Figure 10 - Sample TCP Flags Specification
Seq

This rule option refers to the TCP sequence number.  Essentially, it detects if the packet has a static sequence number set, and is therefore pretty much unused.  It was included for the sake of completeness.

Format:

seq: <number>;
Ack

The ack rule option keyword refers to the TCP header's acknowledge field.  This rule has one practical purpose so far: detecting  NMAP TCP pings.  A NMAP TCP ping sets this field to zero and sends a packet with the TCP ACK flag set to determine if a network host is active.  The rule to detect this activity is shown in Figure 10.

Format:

ack: <number>;

 
alert any any -> 192.168.1.0/24 any (flags: A; ack: 0; msg: "NMAP TCP ping";)
Figure 11 - TCP ACK Field Usage
Itype

This rule tests the value of the ICMP type field.  It is set using the numeric value of this field.  For a list of the available values, look in the decode.h file included with Snort or in any ICMP reference.  It should be noted that the values can be set out of range to detect invalid ICMP type values that are sometimes used in denial of service and flooding attacks.

Format:

itype: <number>;
Icode

The icode rule option keyword is pretty much identical to the itype rule, just set a numeric value in here and Snort will detect any traffic using that ICMP code value.  Out of range values can also be set to detect suspicious traffic.

Format:

icode: <number>;
Session

The session keyword is brand new as of version 1.3.1.1 and is used to extract the user data from TCP sessions.  It is extremely useful for seeing what users are typing in telnet, rlogin, ftp, or even web sessions.  There are two available argument keywords for the session rule option, printable or all.  The printable keyword only prints out data that the user would normally see or be able to type.  The all keyword substitutes non-printable characters with their hexadecimal equivalents.  This function can slow Snort down considerably, so it shouldn't be used in heavy load situations, and is probably best suited for post-processing binary (tcpdump format) log files.  See Figure 11 for a good example of a telnet session logging rule.

Format:

session: [printable|all];

 
log tcp any any <> 192.168.1.0/24 23 (session: printable;)
Figure 12 - Logging Printable Telnet Session Data
Icmp_id

The icmp_id option examines an ICMP ECHO packet's ICMP ID number for a specific value. This is useful because some covert channel programs use static ICMP fields when they communicate. This particular plugin was developed to enable the stacheldraht detection rules written by Max Vision, but it is certainly useful for detection of a number of potential attacks.

Format:

icmp_id: <number>;
Icmp_seq

The icmp_id option examines an ICMP ECHO packet's ICMP sequence field for a specific value. This is useful because some covert channel programs use static ICMP fields when they communicate. This particular plugin was developed to enable the stacheldraht detection rules written by Max Vision, but it is certainly useful for detection of a number of potential attacks. (And yes, I know the info for this field is almost identical to the icmp_id description, it's practically the same damn thing!)

Format:

icmp_seq: <number>;
Ipoption

If IP options are present in a packet, this option will search for a specific option in use, such as source routing. Valid arguments to this option are:

The most frequently watched for IP options are strict and loose source routing which aren't used in any widespread internet applications. Only a single option may be specified per rule.

Format:

ipoption: <option>;
Rpc

This option looks at RPC requests and automatically decodes the application, procedure, and program version, indicating success when all three variables are matched. The format of the option call is "application, procedure, version". Wildcards are valid for both the procedure and version numbers and are indicated with a "*".

Format:

icmp_seq: <number, [number|*], [number|*]>;

 
alert tcp any any -> 192.168.1.0/24 111 (rpc: 100000,*,3; msg:"RPC getport (TCP)";)
alert udp any any -> 192.168.1.0/24 111 (rpc: 100000,*,3; msg:"RPC getport (UDP)";)
alert udp any any -> 192.168.1.0/24 111 (rpc: 100083,*,*; msg:"RPC ttdb";)
alert udp any any -> 192.168.1.0/24 111 (rpc: 100232,10,*; msg:"RPC sadmin";)
Figure 13 - Various RPC Call Alerts
Resp

The resp keyword implements flexible reponse (FlexResp) to traffic that matches a Snort rule.  The FlexResp code allows Snort to actively close offending connections.  The following arguments are valid for this module:

These options can be combined to send multiple responses to the target host.  Multiple arguments are separated by a comma.

Format:

resp: <resp_modifier[, resp_modifier...]>;
alert tcp any any -> 192.168.1.0/24 1524 (flags: S; resp: rst_all; msg: "Root shell backdoor attempt";) 
alert udp any any -> 192.168.1.0/24 31 (resp: icmp_port,icmp_host; msg: "Hacker's Paradise access attempt";) 
Figure 14 - FlexResp Usage Examples
Preprocessors
Preprocessor Overview

Preprocessors were introduced in version 1.5 of Snort.  They allow the functionality of Snort to be extended by allowing users and programmers to drop modular "plugins" into Snort fairly easily.  Preprocessor code is run before the detection engine is called, but after the packet has been decoded.  The packet can be modified or analyzed in an "out of band" manner through this mechanism.

Preprocessors are loaded and configured using the preprocessor keyword.  The format of the preprocessor directive in the Snort rules file is:

preprocessor <name>: <options>
preprocessor minfrag: 128 
Figure 15 - Preprocessor Directive Format Example

Available Preprocessor Modules

Minfrag

The minfrag preprocessor examines fragmented packets for a specified size threshold.  When packets are fragmented, it is generally caused by routers between the source and destination.  Generally speaking, there is no piece of commercial network equipment that fragments packets in sizes smaller than 512 bytes, so we can use this fact to enable traffic to be monitored for tiny fragments that are generally indicative of someone trying to hide their traffic behind fragmentation.

Format:

minfrag: <threshold number>

HTTP Decode

HTTP Decode is used to process HTTP URI strings and convert their data to non-obfuscated ASCII strings.  This is done to defeat evasive web URL scanners and hostile attackers that could otherwise elude the content analysis strings used to examine HTTP traffic for suspicious activity.  The preprocessor module takes HTTP port numbers (separated by spaces) to be normalized as its arguments (typically 80 and 8080).

Format:

http_decode: <port list>
preprocessor http_decode: 80 8080 
Figure 16 - HTTP Decode Directive Format Example
Portscan Detector

The Snort Portscan Preprocessor is developed by Patrick Mullen and (much) more information is available at his web page.

What the Snort Portscan Preprocessor does:

Log the start and end of portscans from a single source IP to the standard logging facility.

If a log file is specified, logs the destination IPs and ports scanned as well as the type of scan.

A portscan is defined as TCP connection attempts to more than P ports in T seconds or UDP packets sent to more than P ports in T seconds.  Ports can be spread across any number of destination IP addresses, and may all be the same port if spread across multiple IPs. This version does single->single and single->many portscans.  The next full release will do distributed portscans (multiple->single or multiple->multiple).  A portscan is also defined as a single "stealth scan" packet, such as NULL, FIN, SYNFIN, XMAS, etc.  This means that from scan-lib in the standard distribution of snort you should comment out the section for stealth scan packets.  The benefit is with the portscan module these alerts would only show once per scan, rather than once for each packet.  If you use the external logging feature you can look at the technique and type in the log file.

The arguments to this module are:

Format:

portscan: <network to monitor> <number of ports> <detection period> <logdir/filename>
preprocessor portscan: 192.168.1.0/24 5 7 /var/log/portscan.log 
Figure 17 - Portscan Module Configuration Example
Portscan Ignorehosts

Another module from Patrick Mullen that modifies the portscan detection system's operation.  If you have servers which tend to trip off the portscan detector (such as NTP, NFS, and DNS servers), you can tell portscan to ignore TCP SYN and UDP portscans from certain hosts.   The arguments to this module are a list of IPs/CIDR blocks to be ignored.

Format:

portscan-ignorehosts: <host list>
preprocessor portscan-ignorehosts: 192.168.1.5/32 192.168.3.0/24 
Figure 18 - Portscan Ignorehosts Module Configuration Example
Output Modules
Output Module Overview

Output modules are new as of version 1.6.  They allow Snort to be much more flexible in the formatting and presentation of output to its users.  The output modules are run when the alert or logging subsystems of Snort are called, after the preprocessors and detection engine. The format of the directives in the rules file is very similar to that of the preprocessors.

NOTE:  Output modules specified in the Snort rules file are overriden if a command line output swith is specified at runtime.  For instance, if the alert_syslog module is specified in the rules file and the "-A fast" command line option is used, the alert_syslog module is disabled and the command line switch is obeyed.

Output modules are loaded at runtime by specifying the output keyword in the rules file:

output <name>: <options>
output alert_syslog: LOG_AUTH LOG_ALERT 
Figure 19 - Output Module Configuration Example

Available Output Modules

Alert_syslog

This module sends alerts to the syslog facility (much like the -s command line switch).  This module also allows the user to specify the logging facility and priority within the Snort rules file, giving users greater flexibility in logging alerts. 

Available keywords:

Format:

alert_syslog: <facility> <priority> <options>

Log_tcpdump

The log_tcpdump module logs packets to a tcpdump-formatted file. This is useful for performing post process analysis on collected traffic with the vast number of tools that are avialable for examining tcpdump formatted files.  This module only takes a single argument, the name of the output file.

Format:

log_tcpdump: <output filename>
output log_tcpdump: snort.log 
Figure 20 - Tcpdump Output Module Configuration Example
Log_postgresql

This module from Jed Pickel logs Snort data to a Postgres SQL database.  More information on installing and configuring this module can be found on the Incident.org web page.  The arguments to this plugin are the name of the database to be logged to and a parameter list.  Parameters are specified with the format parameter = argument.  The following parameters are available:

Format:

log_postgresql: <database name>, <parameter list>
Advanced Rule Concepts
Includes

Versions of Snort after 1.3.1.2 include new rules file parsing functionality developed by Christian Lademann, including two new rules file keywords.  The first of these keywords is include.  The include keyword allows other rule files to be included with the rules file that indicated on the Snort command line.

Format:

include: <include file path/name>
Note that there is no semicolon at the end of this line.  Included files will substitute any predefined variable values into their own variable references.  See the Variables section for more information on defining and using variables in Snort rule files.

Variables

As of version 1.3.1.2, variables may be defined in Snort.  These are simple substitution variables set with the var keyword as in Figure 21.

Format:

var: <name> <value>

var MY_NET 192.168.1.0/24

alert tcp any any -> $MY_NET any (flags: S; msg: "SYN packet";)
 
Figure 21 - Example of Variable Definition and Usage

The rule variable names can be modified in several ways.  You can define meta-variables using the "$" operator.  These can be used with the variable modifier operators, "?" and "-".

See figure 13 for an example of these rules modifiers in action.
 
 

var MY_NET $(MY_NET:-192.168.1.0/24)

log tcp any any -> $(MY_NET:?MY_NET is undefined!) 23
 
Figure 22 - Advanced Variable Usage Example

 


Building Good Rules
There are some general concepts to keep in mind when developing Snort rules to maximize efficiency and speed.  I will add to this section as my muse wills. :)

Content Rules are Case Sensitive (unless you use the "nocase" option)

Don't forget that content rules are case sensitive and that many programs typically use uppercase letters to indicate commands.  FTP is a good example of this.  Consider the following two rules:

alert tcp any any -> 192.168.1.0/24 21 (content: "user root"; msg: "FTP root login";)

alert tcp any any -> 192.168.1.0/24 21 (content: "USER root"; msg: "FTP root login";)

The second of those two rules will catch most every automated root login attempt, but none that use lower case characters for "user". 

Speeding Up Rules That Have Content Options

The order that rules are tested by the detection engine is completely independent of the order that they are written in a rule.  The last rule test that is done (when necessary) is always the content rule option.  Take advantage of this fact by using other faster rule options that can detect whether or not the content needs to be checked at all.  For instance, most of the time when data is sent from client to server after a TCP session is established, the PSH and ACK TCP flags are set on the packet containing the data.  This fact can be taken advantage of by rules that need to test payload content coming from the client to the sever with a simple TCP flag test that is far less computationally expensive than the pattern match algorithm.  Knowing this, a simple way to speed up rules that use content options is to also perform a flag test, as in Figure 23.  The basic idea is that if the PSH and ACK flags aren't set, there's no need to test the packet payload for the given rule.  If the flags are set, the additional computing power required to perform the test is negligible.
 
 

alert tcp any any -> 192.168.1.0/24 80 (content: "cgi-bin/phf"; flags: PA; msg: "CGI-PHF probe";)
Figure 23 - Using TCP Flag Tests to Hasten Content Rules

 



 

Version 1.1, All rights reserved, © Copyright 1999,2000 Martin Roesch