Version 2.5.11 (Nov 4, 2009)
Copyright © 2004-2009 Breach Security, Inc. (http://www.breach.com)
Table of Contents
SecAction
SecArgumentSeparator
SecAuditEngine
SecAuditLog
SecAuditLog2
SecAuditLogDirMode
SecAuditLogFileMode
SecAuditLogParts
SecAuditLogRelevantStatus
SecAuditLogStorageDir
SecAuditLogType
SecCacheTransformations
(Deprecated/Experimental)SecChrootDir
SecComponentSignature
SecContentInjection
SecCookieFormat
SecDataDir
SecDebugLog
SecDebugLogLevel
SecDefaultAction
SecGeoLookupDb
SecGuardianLog
SecMarker
SecPdfProtect
SecPdfProtectMethod
SecPdfProtectSecret
SecPdfProtectTimeout
SecPdfProtectTokenName
SecRequestBodyAccess
SecRequestBodyLimit
SecRequestBodyNoFilesLimit
SecRequestBodyInMemoryLimit
SecResponseBodyLimit
SecResponseBodyLimitAction
SecResponseBodyMimeType
SecResponseBodyMimeTypesClear
SecResponseBodyAccess
SecRule
SecRuleInheritance
SecRuleEngine
SecRuleRemoveById
SecRuleRemoveByMsg
SecRuleScript
(Experimental)SecRuleUpdateActionById
SecServerSignature
SecTmpDir
SecUploadDir
SecUploadFileMode
SecUploadKeepFiles
SecWebAppId
ARGS
ARGS_COMBINED_SIZE
ARGS_NAMES
ARGS_GET
ARGS_GET_NAMES
ARGS_POST
ARGS_POST_NAMES
AUTH_TYPE
ENV
FILES
FILES_COMBINED_SIZE
FILES_NAMES
FILES_SIZES
FILES_TMPNAMES
GEO
HIGHEST_SEVERITY
MATCHED_VAR
MATCHED_VAR_NAME
MODSEC_BUILD
MULTIPART_CRLF_LF_LINES
MULTIPART_STRICT_ERROR
MULTIPART_UNMATCHED_BOUNDARY
PATH_INFO
QUERY_STRING
REMOTE_ADDR
REMOTE_HOST
REMOTE_PORT
REMOTE_USER
REQBODY_PROCESSOR
REQBODY_PROCESSOR_ERROR
REQBODY_PROCESSOR_ERROR_MSG
REQUEST_BASENAME
REQUEST_BODY
REQUEST_COOKIES
REQUEST_COOKIES_NAMES
REQUEST_FILENAME
REQUEST_HEADERS
REQUEST_HEADERS_NAMES
REQUEST_LINE
REQUEST_METHOD
REQUEST_PROTOCOL
REQUEST_URI
REQUEST_URI_RAW
RESPONSE_BODY
RESPONSE_CONTENT_LENGTH
RESPONSE_CONTENT_TYPE
RESPONSE_HEADERS
RESPONSE_HEADERS_NAMES
RESPONSE_PROTOCOL
RESPONSE_STATUS
RULE
SCRIPT_BASENAME
SCRIPT_FILENAME
SCRIPT_GID
SCRIPT_GROUPNAME
SCRIPT_MODE
SCRIPT_UID
SCRIPT_USERNAME
SERVER_ADDR
SERVER_NAME
SERVER_PORT
SESSION
SESSIONID
TIME
TIME_DAY
TIME_EPOCH
TIME_HOUR
TIME_MIN
TIME_MON
TIME_SEC
TIME_WDAY
TIME_YEAR
TX
USERID
WEBAPPID
WEBSERVER_ERROR_LOG
XML
base64Decode
base64Encode
compressWhitespace
escapeSeqDecode
hexDecode
hexEncode
htmlEntityDecode
jsDecode
length
lowercase
md5
none
normalisePath
normalisePathWin
parityEven7bit
parityOdd7bit
parityZero7bit
removeNulls
removeWhitespace
replaceComments
replaceNulls
urlDecode
urlDecodeUni
urlEncode
sha1
trimLeft
trimRight
trim
ModSecurity is a web application firewall (WAF). With over 70% of attacks now carried out over the web application level, organisations need all the help they can get in making their systems secure. WAFs are deployed to establish an increased external security layer to detect and/or prevent attacks before they reach web applications. ModSecurity provides protection from a range of attacks against web applications and allows for HTTP traffic monitoring and real-time analysis with little or no changes to existing infrastructure.
Web servers are typically well-equipped to log traffic in a form useful for marketing analyses, but fall short logging traffic to web applications. In particular, most are not capable of logging the request bodies. Your adversaries know this, and that is why most attacks are now carried out via POST requests, rendering your systems blind. ModSecurity makes full HTTP transaction logging possible, allowing complete requests and responses to be logged. Its logging facilities also allow fine-grained decisions to be made about exactly what is logged and when, ensuring only the relevant data is recorded. As some of the request and/or response may contain sensitive data in certain fields, ModSecurity can be configured to mask these fields before they are written to the audit log.
In addition to providing logging facilities, ModSecurity can monitor the HTTP traffic in real time in order to detect attacks. In this case, ModSecurity operates as a web intrusion detection tool, allowing you to react to suspicious events that take place at your web systems.
ModSecurity can also act immediately to prevent attacks from reaching your web applications. There are three commonly used approaches:
Negative security model. A negative security model monitors requests for anomalies, unusual behaviour, and common web application attacks. It keeps anomaly scores for each request, IP addresses, application sessions, and user accounts. Requests with high anomaly scores are either logged or rejected altogether.
Positive security model. When a positive security model is deployed, only requests that are known to be valid are accepted, with everything else rejected. This model requires knownledge of the web applications you are protecting. Therefore a positive security model works best with applications that are heavily used but rarely updated so that maintenance of the model is minimized.
Known weaknesses and vulnerabilities. Its rule language makes ModSecurity an ideal external patching tool. External patching (sometimes referred to as Virtual Patching) is about reducing the window of opportunity. Time needed to patch application vulnerabilities often runs to weeks in many organisations. With ModSecurity, applications can be patched from the outside, without touching the application source code (and even without any access to it), making your systems secure until a proper patch is applied to the application.
A flexible rule engine sits in the heart of ModSecurity. It implements the ModSecurity Rule Language, which is a specialised programming language designed to work with HTTP transaction data. The ModSecurity Rule Language is designed to be easy to use, yet flexible: common operations are simple while complex operations are possible. Certified ModSecurity Rules, included with ModSecurity, contain a comprehensive set of rules that implement general-purpose hardening, protocol validation and detection of common web application security issues. Heavily commented, these rules can be used as a learning tool.
ModSecurity is an embeddable web application firewall, which means it can be deployed as part of your existing web server infrastructure provided your web servers are Apache-based. This deployment method has certain advantages:
No changes to existing network. It only takes a few minutes to add ModSecurity to your existing web servers. And because it was designed to be completely passive by default, you are free to deploy it incrementally and only use the features you need. It is equally easy to remove or deactivate it if required.
No single point of failure. Unlike with network-based deployments, you will not be introducing a new point of failure to your system.
Implicit load balancing and scaling. Because it works embedded in web servers, ModSecurity will automatically take advantage of the additional load balancing and scalability features. You will not need to think of load balancing and scaling unless your existing system needs them.
Minimal overhead. Because it works from inside the web server process there is no overhead for network communication and minimal overhead in parsing and data exchange.
No problem with encrypted or compressed content. Many IDS systems have difficulties analysing SSL traffic. This is not a problem for ModSecurity because it is positioned to work when the traffic is decrypted and decompressed.
ModSecurity works equally well when deployed as part of an Apache-based reverse proxy server, and many of our customers choose to do so. In this scenario, one installation of ModSecurity can protect any number of web servers (even the non-Apache ones).
ModSecurity is known to work well on a wide range of operating systems. Our customers are successfully running it on Linux, Windows, Solaris, FreeBSD, OpenBSD, NetBSD, AIX, Mac OS X, and HP-UX.
ModSecurity is available under two licenses. Users can choose to use the software under the terms of the GNU General Public License version 2 (licence text is included with the distribution), as an Open Source / Free Software product. A range of commercial licenses is also available, together with a range of commercial support contracts. For more information on commercial licensing please contact Breach Security.
ModSecurity, mod_security, ModSecurity Pro, and ModSecurity Core Rules are trademarks or registered trademarks of Breach Security, Inc.
ModSecurity is a web application firewall engine that provides very little protection on its own. In order to become useful, ModSecurity must be configured with rules. In order to enable users to take full advantage of ModSecurity out of the box, Breach Security, Inc. is providing a free certified rule set for ModSecurity 2.x. Unlike intrusion detection and prevention systems, which rely on signatures specific to known vulnerabilities, the Core Rules provide generic protection from unknown vulnerabilities often found in web applications, which are in most cases custom coded. The Core Rules are heavily commented to allow it to be used as a step-by-step deployment guide for ModSecurity. The latest Core Rules can be found at the ModSecurity website - http://www.modsecurity.org/projects/rules/.
In order to provide generic web applications protection, the Core Rules use the following techniques:
HTTP protection - detecting violations of the HTTP protocol and a locally defined usage policy.
Common Web Attacks Protection - detecting common web application security attack.
Automation detection - Detecting bots, crawlers, scanners and other surface malicious activity.
Trojan Protection - Detecting access to Trojans horses.
Error Hiding - Disguising error messages sent by the server.
ModSecurity installation requirements:
ModSecurity 2.x works only with Apache 2.0.x or higher. Version 2.2.x is highly recommended.
Make sure you have mod_unique_id
installed.
mod_unique_id is packaged with Apache httpd.
libapr and libapr-util
libpcre
libxml2
liblua v5.1.x
This library is optional and only needed if you will be using the new Lua engine.
http://www.lua.org/download.html
Note that ModSecurity requires the dynamic libraries. These are not built by default in the source distribution, so the binary distribution is recommended.
libcurl v7.15.1 or higher
If you will be using the ModSecurity Log Collector (mlogc) to send audit logs to a central repository, then you will also need the curl library.
Many have had issues with libcurl linked with the GnuTLS library for SSL/TLS support. It is recommended that the openssl library be used for SSL/TLS support in libcurl.
ModSecurity installation consists of the following steps:
Stop Apache httpd
Unpack the ModSecurity archive
Building differs for UNIX (or UNIX-like) operating systems and Windows.
UNIX
Run the configure script to generate a Makefile. Typically no options are needed.
./configure
Options are available for more customization (use
./configure --help
for a full list), but
typically you will only need to specify the location of the
apxs
command installed by Apache httpd with
the --with-apxs
option.
./configure
--with-apxs=/path/to/httpd-2.x.y/bin/apxs
There are certain configure options that are meant for
debugging an other development use. If enabled, these
options can substantially impact performance. These options
include all --debug-*
options as well as
the --enable-performance-measurements
options.
Compile with: make
Optionally test with: make
test
This is step is still a bit experimental. If you have problems, please send the full output and error from the build to the support list. Most common issues are related to not finding the required headers and/or libraries.
Optionally build the ModSecurity Log Collector with:
make mlogc
Optionally install mlogc
: Review the
INSTALL
file included in the
apache2/mlogc-src directory in the distribution.
Install the ModSecurity module with: make
install
Windows (MS VC++ 8)
Edit Makefile.win
to configure the
Apache base and library paths.
Compile with: nmake -f
Makefile.win
Install the ModSecurity module with: nmake -f
Makefile.win install
Copy the libxml2.dll
and
lua5.1.dll
to the Apache
bin
directory. Alternatively you can follow
the step below for using LoadFile to load these
libraries.
Edit the main Apache httpd config file (usually
httpd.conf
)
On UNIX (and Windows if you did not copy the DLLs as stated above) you must load libxml2 and lua5.1 before ModSecurity with something like this:
LoadFile /usr/lib/libxml2.so LoadFile /usr/lib/liblua5.1.so
Load the ModSecurity module with:
LoadModule security2_module modules/mod_security2.so
Configure ModSecurity
Start Apache httpd
You should now have ModSecurity 2.x up and running.
If you have compiled Apache yourself you might experience problems compiling ModSecurity against PCRE. This is because Apache bundles PCRE but this library is also typically provided by the operating system. I would expect most (all) vendor-packaged Apache distributions to be configured to use an external PCRE library (so this should not be a problem).
You want to avoid Apache using the bundled PCRE library and
ModSecurity linking against the one provided by the operating system.
The easiest way to do this is to compile Apache against the PCRE library
provided by the operating system (or you can compile it against the
latest PCRE version you downloaded from the main PCRE distribution
site). You can do this at configure time using the --with-pcre
switch. If you are not in a
position to recompile Apache, then, to compile ModSecurity successfully,
you'd still need to have access to the bundled PCRE headers (they are
available only in the Apache source code) and change the include path
for ModSecurity (as you did in step 7 above) to point to them (via the
--with-pcre
ModSecurity configure option).
Do note that if your Apache is using an external PCRE library you
can compile ModSecurity with WITH_PCRE_STUDY
defined,which would possibly
give you a slight performance edge in regular expression
processing.
Non-gcc compilers may have problems running out-of-the-box as the current build system was designed around the gcc compiler and some compiler/linker flags may differ. To use a non-gcc compiler you may need some manual Makefile tweaks if issues cannot be solved by exporting custom CFLAGS and CPPFLAGS environment variables.
If you are upgrading from ModSecurity 1.x, please refer to the migration matrix at http://www.modsecurity.org/documentation/ModSecurity-Migration-Matrix.pdf
The following section outlines all of the ModSecurity directives.
Most of the ModSecurity directives can be used inside the various Apache
Scope Directives such as VirtualHost
,
Location
, LocationMatch
,
Directory
, etc... There are others, however, that can
only be used once in the main configuration file. This information is
specified in the Scope sections below. The first version to use a given
directive is given in the Version sections below.
These rules, along with the Core rules files, should be contained is
files outside of the httpd.conf file and called up with Apache "Include"
directives. This allows for easier updating/migration of the rules. If you
create your own custom rules that you would like to use with the Core
rules, you should create a file called -
modsecurity_crs_15_customrules.conf
and place it in
the same directory as the Core rules files. By using this file name, your
custom rules will be called up after the standard ModSecurity Core rules
configuration file but before the other Core rules. This allows your rules
to be evaluated first which can be useful if you need to implement
specific "allow" rules or to correct any false positives in the Core rules
as they are applied to your site.
It is highly encouraged that you do not edit the Core rules files
themselves but rather place all changes (such as
SecRuleRemoveByID
, etc...) in your custom rules file.
This will allow for easier upgrading as newer Core rules are released by
Breach Security on the ModSecurity website.
Description: Unconditionally processes the
action list it receives as the first and only parameter. It accepts one
parameter, the syntax of which is identical to the third parameter
of SecRule
.
Syntax: SecAction
action1,action2,action3
Example Usage: SecAction
nolog,phase:1,initcol:RESOURCE=%{REQUEST_FILENAME}
Processing Phase: Any
Scope: Any
Version: 2.0.0
Dependencies/Notes: None
SecAction is best used when you unconditionally execute an action.
This is explicit triggering whereas the normal Actions are conditional
based on data inspection of the request/response. This is a useful
directive when you want to run certain actions such as
initcol
to initialize collections.
Description: Specifies which character to use
as separator for
application/x-www-form-urlencoded
content. Defaults to
&
. Applications are sometimes
(very rarely) written to use a semicolon (;
).
Syntax: SecArgumentSeparator character
Example Usage: SecArgumentSeparator ;
Processing Phase: Any
Scope: Main
Version: 2.0.0
Dependencies/Notes: None
This directive is needed if a backend web application is using a non-standard argument separator. If this directive is not set properly for each web application, then ModSecurity will not be able to parse the arguments appropriately and the effectiveness of the rule matching will be significantly decreased.
Description: Configures the audit logging engine.
Syntax: SecAuditEngine On|Off|RelevantOnly
Example Usage: SecAuditEngine On
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: Can be set/changed with
the "ctl
" action for the current transaction.
Example: The following example shows the various audit directives used together.
SecAuditEngine RelevantOnly SecAuditLog logs/audit/audit.log SecAuditLogParts ABCFHZ SecAuditLogType concurrent SecAuditLogStorageDir logs/audit SecAuditLogRelevantStatus ^(?:5|4\d[^4])
Possible values are:
On
- log all transactions
by default.
Off
- do not log
transactions by default.
RelevantOnly
- by default
only log transactions that have triggered a warning or an error, or
have a status code that is considered to be relevant (see SecAuditLogRelevantStatus
).
Description: Defines the path to the main audit log file.
Syntax: SecAuditLog
/path/to/auditlog
Example Usage: SecAuditLog
/usr/local/apache/logs/audit.log
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: This file is open on startup when the server typically still runs as root. You should not allow non-root users to have write privileges for this file or for the directory it is stored in..
This file will be used to store the audit log entries if serial audit logging format is used. If concurrent audit logging format is used this file will be used as an index, and contain a record of all audit log files created. If you are planning to use Concurrent audit logging and sending your audit log data off to a remote Console host or commercial ModSecurity Management Appliance, then you will need to configure and use the ModSecurity Log Collector (mlogc) and use the following format for the audit log:
SecAuditLog "|/path/to/mlogc /path/to/mlogc.conf"
Description: Defines the path to the
secondary audit log index file when concurrent logging is enabled. See
SecAuditLog2
for more details.
Syntax: SecAuditLog2
/path/to/auditlog2
Example Usage: SecAuditLog2
/usr/local/apache/logs/audit2.log
Processing Phase: N/A
Scope: Any
Version: 2.1.2
Dependencies/Notes: A main audit log must be
defined via SecAuditLog
before this
directive may be used. Additionally, this log is only used for
replicating the main audit log index file when concurrent audit logging
is used. It will not be used for non-concurrent
audit logging.
Description: Configures the mode
(permissions) of any directories created for concurrent audit logs using
an octal mode (as used in chmod). See SecAuditLogFileMode
for controlling the mode
of audit log files.
Syntax: SecAuditLogDirMode octal_mode|"default"
Example Usage: SecAuditLogDirMode 02750
Processing Phase: N/A
Scope: Any
Version: 2.5.10
Dependencies/Notes: This feature is not available on operating systems not supporting octal file modes. The default mode (0600) only grants read/write access to the account writing the file. If access from another account is needed (using mpm-itk is a good example), then this directive may be required. However, use this directive with caution to avoid exposing potentially sensitive data to unauthorized users. Using the value "default" will revert back to the default setting.
The process umask may still limit the mode if it is being more restrictive than the mode set using this directive.
Description: Configures the mode
(permissions) of any files created for concurrent audit logs using an
octal mode (as used in chmod). See SecAuditLogDirMode
for controlling the mode of
created audit log directories.
Syntax: SecAuditLogFileMode
octal_mode|"default"
Example Usage: SecAuditLogFileMode 00640
Processing Phase: N/A
Scope: Any
Version: 2.5.10
Dependencies/Notes: This feature is not available on operating systems not supporting octal file modes. The default mode (0600) only grants read/write access to the account writing the file. If access from another account is needed (using mpm-itk is a good example), then this directive may be required. However, use this directive with caution to avoid exposing potentially sensitive data to unauthorized users. Using the value "default" will revert back to the default setting.
The process umask may still limit the mode if it is being more restrictive than the mode set using this directive.
Description: Defines which part of each transaction are going to be recorded in audit log. Each part is assigned a single letter. If a letter appears in the list then the equivalent part of each transactions will be recorded. See below for the list of all parts.
Syntax: SecAuditLogParts PARTS
Example Usage: SecAuditLogParts ABCFHZ
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: At this time ModSecurity
does not log response bodies of stock Apache responses (e.g. 404
), or the Server
and Date
response headers.
Default: ABCFHZ
.
Please refer to the ModSecurity Data Formats document for a detailed description of every available part.
Available audit log parts:
A
- audit log header
(mandatory)
B
- request headers
C
- request body (present
only if the request body exists and ModSecurity is configured to
intercept it)
D
- RESERVED for
intermediary response headers, not implemented yet.
E
- intermediary response
body (present only if ModSecurity is configured to intercept
response bodies, and if the audit log engine is configured to record
it). Intermediary response body is the same as the actual response
body unless ModSecurity intercepts the intermediary response body,
in which case the actual response body will contain the error
message (either the Apache default error message, or the
ErrorDocument page).
F
- final response headers
(excluding the Date and Server headers, which are always added by
Apache in the late stage of content delivery).
G
- RESERVED for the actual
response body, not implemented yet.
H
- audit log
trailer
I
- This part is a
replacement for part C. It will log the same data as C in all cases
except when multipart/form-data
encoding in used. In this case it will log a fake application/x-www-form-urlencoded
body
that contains the information about parameters but not about the
files. This is handy if you don't want to have (often large) files
stored in your audit logs.
J
- RESERVED. This part,
when implemented, will contain information about the files uploaded
using multipart/form-data
encoding.
K
- This part contains a
full list of every rule that matched (one per line) in the order
they were matched. The rules are fully qualified and will thus show
inherited actions and default operators. Supported as of
v2.5.0
Z
- final boundary,
signifies the end of the entry (mandatory)
Description: Configures which response status code is to be considered relevant for the purpose of audit logging.
Syntax: SecAuditLogRelevantStatus REGEX
Example Usage: SecAuditLogRelevantStatus
^(?:5|4\d[^4])
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: Must have the
SecAuditEngine
set to
RelevantOnly
. The parameter is a regular
expression.
The main purpose of this directive is to allow you to configure audit logging for only transactions that generate the specified HTTP Response Status Code. This directive is often used to the decrease the total size of the audit log file. Keep in mind that if this parameter is used, then successful attacks that result in a 200 OK status code will not be logged.
Description: Configures the storage directory where concurrent audit log entries are to be stored.
Syntax: SecAuditLogStorageDir
/path/to/storage/dir
Example Usage: SecAuditLogStorageDir
/usr/local/apache/logs/audit
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: SecAuditLogType must be set to Concurrent. The directory must already be created before starting Apache and it must be writable by the web server user as new files are generated at runtime.
As with all logging mechanisms, ensure that you specify a file system location that has adequate disk space and is not on the root partition.
Description: Configures the type of audit logging mechanism to be used.
Syntax: SecAuditLogType Serial|Concurrent
Example Usage: SecAuditLogType Serial
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: Must specify
SecAuditLogStorageDir
if you use concurrent
logging.
Possible values are:
Serial
- all audit log
entries will be stored in the main audit logging file. This is more
convenient for casual use but it is slower as only one audit log
entry can be written to the file at any one file.
Concurrent
- audit log
entries will be stored in separate files, one for each transaction.
Concurrent logging is the mode to use if you are going to send the
audit log data off to a remote ModSecurity Console host.
Description: Controls caching of transformations. Caching is off by default starting with 2.5.6, when it was deprecated and downgraded back to experimental.
Syntax: SecCacheTransformations On|Off
[options]
Example Usage: SecCacheTransformations On
"minlen:64,maxlen:0"
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: N/A
First parameter:
On
- cache transformations
(per transaction, per phase) allowing identical transformations to
be performed only once. (default)
Off
- do not cache any
transformations, forcing all transformations to be performed for
each rule executed.
The following options are allowed (comma separated):
incremental:on|off
-
enabling this option will cache every transformation instead of just
the final transformation. (default: off)
maxitems:N
- do not allow
more than N transformations to be cached. The cache will then be
disabled. A zero value is interpreted as "unlimited". This option
may be useful to limit caching for a form with a large number of
ARGS. (default: 512)
minlen:N
- do not cache the
transformation if the value's length is less than N bytes. (default:
32)
maxlen:N
- do not cache the
transformation if the value's length is more than N bytes. A zero
value is interpreted as "unlimited". (default: 1024)
Description: Configures the directory path that will be used to jail the web server process.
Syntax: SecChrootDir
/path/to/chroot/dir
Example Usage: SecChrootDir /chroot
Processing Phase: N/A
Scope: Main
Version: 2.0.0
Dependencies/Notes: This feature is not available on Windows builds. The internal chroot functionality provided by ModSecurity works great for simple setups. One example of a simple setup is Apache serving static files only, or running scripts using modules.builds. Some problems you might encounter with more complex setups:
DNS lookups do not work (this is because this feature requires a shared library that is loaded on demand, after chroot takes place).
You cannot send email from PHP because it uses sendmail and sendmail is outside the jail.
In some cases Apache graceful (reload) no longer works.
You should be aware that the internal chroot feature might not be
100% reliable. Due to the large number of default and third-party
modules available for the Apache web server, it is not possible to
verify the internal chroot works reliably with all of them. A module,
working from within Apache, can do things that make it easy to break out
of the jail. In particular, if you are using any of the modules that
fork in the module initialisation phase (e.g.
mod_fastcgi
, mod_fcgid
,
mod_cgid
), you are advised to examine each Apache
process and observe its current working directory, process root, and the
list of open files. Consider what your options are and make your own
decision.
Description: Appends component signature to the ModSecurity signature.
Syntax: SecComponentSignature
"COMPONENT_NAME/X.Y.Z (COMMENT)"
Example usage: SecComponentSignature
"Core Rules/1.2.3"
Processing Phase: N/A
Scope: Main
Version: 2.5.0
Dependencies/Notes: This directive should be used to make the presence of significant ModSecurity components known. The entire signature will be recorded in transaction audit log. It should be used by ModSecurity module and rule set writers to make debugging easier.
Description: Enables content injection using
actions append
and prepend
.
Syntax: SecContentInjection
(On|Off)
Example Usage: SecContentInjection
On
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: N/A
Description: Selects the cookie format that will be used in the current configuration context.
Syntax: SecCookieFormat 0|1
Example Usage: SecCookieFormat 0
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: None
Possible values are:
0
- use version 0
(Netscape) cookies. This is what most applications use. It is the
default value.
1
- use version 1
cookies.
Description: Path where persistent data (e.g. IP address data, session data, etc) is to be stored.
Syntax: SecDataDir
/path/to/dir
Example Usage: SecDataDir /usr/local/apache/logs/data
Processing Phase: N/A
Scope: Main
Dependencies/Notes: This directive is needed when initcol, setsid an setuid are used. Must be writable by the web server user.
Description: Path to the ModSecurity debug log file.
Syntax: SecDebugLog
/path/to/modsec-debug.log
Example Usage: SecDebugLog
/usr/local/apache/logs/modsec-debug.log
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: None
Description: Configures the verboseness of the debug log data.
Syntax: SecDebugLogLevel 0|1|2|3|4|5|6|7|8|9
Example Usage: SecDebugLogLevel 4
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: Levels 1 - 3
are always sent to the Apache error log.
Therefore you can always use level 0
as the default logging level in production. Level 5
is useful when debugging. It is not
advisable to use higher logging levels in production as excessive
logging can slow down server significantly.
Possible values are:
0
- no logging.
1
- errors (intercepted
requests) only.
2
- warnings.
3
- notices.
4
- details of how
transactions are handled.
5
- as above, but including
information about each piece of information handled.
9
- log everything,
including very detailed debugging information.
Description: Defines the default action to take on a rule match.
Syntax: SecDefaultAction
action1,action2,action3
Example Usage: SecDefaultAction
log,auditlog,deny,status:403,phase:2
Processing Phase: Any
Scope: Any
Version: 2.0.0
Dependencies/Notes: Rules following a
SecDefaultAction
directive will inherit this setting
unless a specific action is specified for an individual rule or until
another SecDefaultAction
is specified. Take special
note that in the logging disruptive actions are not allowed, but this
can inadvertently be inherited using a disruptive action in
SecDefaultAction
.
The default value is minimal (differing from previous versions):
SecDefaultAction phase:2,log,auditlog,pass
SecDefaultAction
must specify a disruptive
action and a processing phase and cannot contain metadata
actions.
SecDefaultAction
is not
inherited across configuration contexts. (For an example of why this
may be a problem for you, read the following ModSecurity Blog entry
http://blog.modsecurity.org/2008/07/modsecurity-tri.html).
Description: Defines the path to the geographical database file.
Syntax: SecGeoLookupDb /path/to/db
Example Usage: SecGeoLookupDb
/usr/local/geo/data/GeoLiteCity.dat
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: Check out
maxmind.com
for free database files.
Description: Configuration directive to use the httpd-guardian script to monitor for Denial of Service (DoS) attacks.
Syntax: SecGuardianLog |/path/to/httpd-guardian
Example Usage: SecGuardianLog
|/usr/local/apache/bin/httpd-guardian
Processing Phase: N/A
Scope: Main
Version: 2.0.0
Dependencies/Notes: By default httpd-guardian will defend against clients that send more than 120 requests in a minute, or more than 360 requests in five minutes.
Since 1.9, ModSecurity supports a new directive, SecGuardianLog, that is designed to send all access data to another program using the piped logging feature. Since Apache is typically deployed in a multi-process fashion, making information sharing difficult, the idea is to deploy a single external process to observe all requests in a stateful manner, providing additional protection.
Development of a state of the art external protection tool will be a focus of subsequent ModSecurity releases. However, a fully functional tool is already available as part of the Apache httpd tools project. The tool is called httpd-guardian and can be used to defend against Denial of Service attacks. It uses the blacklist tool (from the same project) to interact with an iptables-based (Linux) or pf-based (*BSD) firewall, dynamically blacklisting the offending IP addresses. It can also interact with SnortSam (http://www.snortsam.net). Assuming httpd-guardian is already configured (look into the source code for the detailed instructions) you only need to add one line to your Apache configuration to deploy it:
SecGuardianLog |/path/to/httpd-guardian
Description: Adds a fixed rule marker in the
ruleset to be used as a target in a skipAfter
action.
A SecMarker
directive essentially creates a rule that
does nothing and whose only purpose it to carry the given ID.
Syntax: SecMarker
ID
Example Usage: SecMarker 9999
Processing Phase: Any
Scope: Any
Version: 2.5.0
Dependencies/Notes: None
SecRule REQUEST_URI "^/$" \ "chain,t:none,t:urlDecode,t:lowercase,t:normalisePath,skipAfter:99" SecRule REMOTE_ADDR "^127\.0\.0\.1$" "chain" SecRule REQUEST_HEADERS:User-Agent \ "^Apache \(internal dummy connection\)$" "t:none" SecRule &REQUEST_HEADERS:Host "@eq 0" \ "deny,log,status:400,id:08,severity:4,msg:'Missing a Host Header'" SecRule &REQUEST_HEADERS:Accept "@eq 0" \ "log,deny,log,status:400,id:15,msg:'Request Missing an Accept Header'" SecMarker 99
Description: Enables the PDF XSS protection
functionality. Once enabled access to PDF files is tracked. Direct
access attempts are redirected to links that contain one-time tokens.
Requests with valid tokens are allowed through unmodified. Requests with
invalid tokens are also allowed through but with forced download of the
PDF files. This implementation uses response headers to detect PDF files
and thus can be used with dynamically generated PDF files that do not
have the .pdf
extension in the request URI.
Syntax: SecPdfProtect On|Off
Example Usage: SecPdfProtect On
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: None
Description: Configure desired protection
method to be used when requests for PDF files are detected. Possible
values are TokenRedirection
and
ForcedDownload
. The token redirection approach will
attempt to redirect with tokens where possible. This allows PDF files to
continue to be opened inline but only works for GET requests. Forced
download always causes PDF files to be delivered as opaque binaries and
attachments. The latter will always be used for non-GET requests. Forced
download is considered to be more secure but may cause usability
problems for users ("This PDF won't open anymore!").
Syntax: SecPdfProtectMethod method
Example Usage: SecPdfProtectMethod TokenRedirection
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: None
Default:
TokenRedirection
Description: Defines the secret that will be used to construct one-time tokens. You should use a reasonably long value for the secret (e.g. 16 characters is good). Once selected the secret should not be changed as it will break the tokens that were sent prior to change. But it's not a big deal even if you change it. It will just force download of PDF files with tokens that were issued in the last few seconds.
Syntax: SecPdfProtectSecret secret
Example Usage: SecPdfProtectSecret
MyRandomSecretString
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: None
Description: Defines the token timeout. After token expires it can no longer be used to allow access to PDF file. Request will be allowed through but the PDF will be delivered as attachment.
Syntax: SecPdfProtectTimeout timeout
Example Usage: SecPdfProtectTimeout 10
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: None
Default: 10
Description: Defines the name of the token. The only reason you would want to change the name of the token is if you wanted to hide the fact you are running ModSecurity. It's a good reason but it won't really help as the adversary can look into the algorithm used for PDF protection and figure it out anyway. It does raise the bar slightly so go ahead if you want to.
Syntax: SecPdfProtectTokenName name
Example Usage: SecPdfProtectTokenName PDFTOKEN
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: None
Default: PDFTOKEN
Description: Configures whether request bodies will be buffered and processed by ModSecurity by default.
Syntax: SecRequestBodyAccess On|Off
Example Usage: SecRequestBodyAccess On
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: This directive is
required if you plan to inspect POST_PAYLOAD
. This
directive must be used along with the "phase:2" processing phase action
and REQUEST_BODY
variable/location. If any of these 3
parts are not configured, you will not be able to inspect the request
bodies.
Possible values are:
On
- access request
bodies.
Off
- do not attempt to
access request bodies.
Description: Configures the maximum request body size ModSecurity will accept for buffering.
Syntax: SecRequestBodyLimit NUMBER_IN_BYTES
Example Usage: SecRequestBodyLimit 134217728
Scope: Any
Version: 2.0.0
Dependencies/Notes: 131072 KB (134217728 bytes) is the default setting. Anything over this limit will be rejected with status code 413 Request Entity Too Large. There is a hard limit of 1 GB.
Description: Configures the maximum request
body size ModSecurity will accept for buffering, excluding the size of
files being transported in the request. This directive comes handy to
further reduce susceptibility to DoS attacks when someone is sending
request bodies of very large sizes. Web applications that require file
uploads must configure SecRequestBodyLimit
to a high
value. Since large files are streamed to disk file uploads will not
increase memory consumption. However, it's still possible for someone to
take advantage of a large request body limit and send non-upload
requests with large body sizes. This directive eliminates that
loophole.
Syntax: SecRequestBodyNoFilesLimit
NUMBER_IN_BYTES
Example Usage: SecRequestBodyLimit 131072
Scope: Any
Version: 2.5.0
Dependencies/Notes: 1 MB (1048576 bytes) is
the default setting. This value is very conservative. For most
applications you should be able to reduce it down to 128 KB or lower.
Anything over the limit will be rejected with status code 413
Request Entity Too Large
. There is a hard limit of 1
GB.
Description: Configures the maximum request body size ModSecurity will store in memory.
Syntax: SecRequestBodyInMemoryLimit
NUMBER_IN_BYTES
Example Usage: SecRequestBodyInMemoryLimit 131072
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: None
By default the limit is 128 KB:
# Store up to 128 KB in memory SecRequestBodyInMemoryLimit 131072
Description: Configures the maximum response body size that will be accepted for buffering.
Syntax: SecResponseBodyLimit NUMBER_IN_BYTES
Example Usage: SecResponseBodyLimit 524228
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: Anything over this limit will be rejected with status code 500 Internal Server Error. This setting will not affect the responses with MIME types that are not marked for buffering. There is a hard limit of 1 GB.
By default this limit is configured to 512 KB:
# Buffer response bodies of up to 512 KB in length SecResponseBodyLimit 524288
Description: Controls what happens once a
response body limit, configured with
SecResponseBodyLimit
, is encountered. By default
ModSecurity will reject a response body that is longer than specified.
Some web sites, however, will produce very long responses making it
difficult to come up with a reasonable limit. Such sites would have to
raise the limit significantly to function properly defying the purpose
of having the limit in the first place (to control memory consumption).
With the ability to choose what happens once a limit is reached site
administrators can choose to inspect only the first part of the
response, the part that can fit into the desired limit, and let the rest
through. Some could argue that allowing parts of responses to go
uninspected is a weakness. This is true in theory but only applies to
cases where the attacker controls the output (e.g. can make it arbitrary
long). In such cases, however, it is not possible to prevent leakage
anyway. The attacker could compress, obfuscate, or even encrypt data
before it is sent back, and therefore bypass any monitoring
device.
Syntax: SecResponseBodyLimitAction
Reject|ProcessPartial
Example Usage:
SecResponseBodyLimitAction ProcessPartial
Processing Phase: N/A
Scope: Any
Version: 2.5.0
Dependencies/Notes: None
Description: Configures which MIME
types are to be considered for response
body buffering.
Syntax: SecResponseBodyMimeType mime/type
Example Usage: SecResponseBodyMimeType text/plain
text/html
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: Multiple SecResponseBodyMimeType
directives can be
used to add MIME
types.
The default value is text/plaintext/html
:
SecResponseBodyMimeType text/plain text/html
Description: Clears the list of MIME
types considered for response body
buffering, allowing you to start populating the list from
scratch.
Syntax: SecResponseBodyMimeTypesClear
Example Usage: SecResponseBodyMimeTypesClear
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: None
Description: Configures whether response bodies are to be buffer and analysed or not.
Syntax: SecResponseBodyAccess On|Off
Example Usage: SecResponseBodyAccess On
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: This directive is required if you plan to inspect HTML responses. This directive must be used along with the "phase:4" processing phase action and RESPONSE_BODY variable/location. If any of these 3 parts are not configured, you will not be able to inspect the response bodies.
Possible values are:
On
- access response bodies
(but only if the MIME type matches, see above).
Off
- do not attempt to
access response bodies.
Description: SecRule
is the main ModSecurity directive. It
is used to analyse data and perform actions based on the results.
Syntax: SecRule
VARIABLES OPERATOR [ACTIONS]
Example Usage: SecRule REQUEST_URI "attack" \
"phase:1,t:none,t:urlDecode,t:lowercase,t:normalisePath"
Processing Phase: Any
Scope: Any
Version: 2.0.0
Dependencies/Notes: None
In general, the format of this rule is as follows:
SecRule VARIABLES OPERATOR [ACTIONS]
The second part, OPERATOR
,
specifies how they are going to be checked. The third (optional) part,
ACTIONS
, specifies what to do
whenever the operator used performs a successful match against a
variable.
The first part, VARIABLES
,
specifies which variables are to be checked. For example, the
following rule will reject a transaction that has the word
dirty in the URI:
SecRule ARGS dirty
Each rule can specify one or more variables:
SecRule ARGS|REQUEST_HEADERS:User-Agent dirty
There is a third format supported by the selection operator - XPath expression. XPath expressions can only used against the special variable XML, which is available only of the request body was processed as XML.
SecRule XML:/xPath/Expression dirty
Not all collections support all selection operator format types. You should refer to the documentation of each collection to determine what is and isn't supported.
A variable can contain one or many pieces of data, depending on the nature of the variable and the way it is used. We've seen examples of both approaches in the previous section. When a variable can contain more than one value we refer to it as a collection.
Collections are always expanded before a rule is run. For example, the following rule:
SecRule ARGS dirty
will be expanded to:
SecRule ARGS:p dirty SecRule ARGS:q dirty
in a requests that has only two parameters, named
p
and q
.
Collections come in several flavours:
Created at runtime using transaction data. For example:
ARGS
(contains a list of all request
parameter values) and REQUEST_HEADERS
(contains a list of all request header values).
The TX
collection is created (empty)
for every transaction. Rules can read from it and write to it
(using the setvar
action, for example), but
the information stored in this collection will not survive the
end of transaction.
There are several collections that can be written to, but
which are persisted to the storage backend. These collections
are used to track clients across transactions. Examples of
collections that fall into this type are IP
,
SESSION
and USER
.
In the simplest possible case you will use a regular expression
pattern as the second rule parameter. This is what we've done in the
examples above. If you do this ModSecurity assumes you want to use the
rx
(regular expression) operator.
You can also explicitly specify the operator you want to use by using
@
, followed by the name of an
operator, at the beginning of the second SecRule
parameter:
SecRule ARGS "@rx dirty"
Note how we had to use double quotes to delimit the second rule parameter. This is because the second parameter now has whitespace in it. Any number of whitespace characters can follow the name of the operator. If there are any non-whitespace characters there, they will all be treated as a special parameter to the operator. In the case of the regular expression operator the special parameter is the pattern that will be used for comparison.
The @ can be the second character if you are using negation to negate the result returned by the operator:
SecRule &ARGS "!@rx ^0$"
Operator results can be negated by using an exclamation mark at
the beginning of the second parameter. The following rule matches if
the word dirty
does not appear
in the User-Agent
request header:
SecRule REQUEST_HEADERS:User-Agent !dirty
You can use the exclamation mark in combination with any parameter. If you do, the exclamation mark needs to go first, followed by the explicit operator reference. The following rule has the same effect as the previous example:
SecRule REQUEST_HEADERS:User-Agent "!@rx dirty"
If you need to use negation in a rule that is going to be applied to several variables then it may not be immediately clear what will happen. Consider the following example:
SecRule ARGS:p|ARGS:q !dirty
The above rule is identical to:
SecRule ARGS:p !dirty SecRule ARGS:q !dirty
Negation is applied to operations against individual operations, not agains the entire variable list.
The third parameter, ACTIONS
,
can be omitted only because there is a helper feature that specifies
the default action list. If the parameter isn't omitted the actions
specified in the parameter will be merged with the default action list
to create the actual list of actions that will be processed on a rule
match.
Description: Configures whether the current context will inherit rules from the parent context (configuration options are inherited in most cases - you should look up the documentation for every directive to determine if it is inherited or not).
Syntax: SecRuleInheritance On|Off
Example Usage: SecRuleInheritance Off
Processing Phase: Any
Scope: Any
Version: 2.0.0
Dependencies/Notes: Resource-specific
contexts (e.g. Location
, Directory
, etc) cannot override
phase1 rules configured in the main server or in
the virtual server. This is because phase 1 is run early in the request
processing process, before Apache maps request to resource. Virtual host
context can override phase 1 rules configured in the main server.
Example: The following example shows where ModSecurity may be enabled in the main Apache configuration scope, however you might want to configure your VirtualHosts differently. In the first example, the first VirtualHost is not inheriting the ModSecurity main config directives and in the second one it is.
SecRuleEngine On SecDefaultAction log,pass,phase:2 ... <VirtualHost *:80> ServerName app1.com ServerAlias www.app1.com SecRuleInheritance Off SecDefaultAction log,deny,phase:1,redirect:http://www.site2.com ... </VirtualHost> <VirtualHost *:80> ServerName app2.com ServerAlias www.app2.com SecRuleInheritance On SecRule ARGS "attack" ... </VirtualHost>
Possible values are:
On
- inherit rules from the
parent context.
Off
- do not inherit rules
from the parent context.
Configuration contexts are an Apache concept. Directives
<Directory>
,
<Files>
,
<Location>
and
<VirtualHost>
are all used to create
configuration contexts. For more information please go to the
Apache documentation section Configuration
Sections.
Description: Configures the rules engine.
Syntax: SecRuleEngine On|Off|DetectionOnly
Example Usage: SecRuleEngine On
Processing Phase: Any
Scope: Any
Version: 2.0.0
Dependencies/Notes: This directive can also be controlled by the ctl action (ctl:ruleEngine=off) for per rule processing.
Possible values are:
On
- process rules.
Off
- do not process
rules.
DetectionOnly
- process
rules but never intercept transactions, even when rules are
configured to do so.
Description: Removes matching rules from the parent contexts.
Syntax: SecRuleUpdateActionById RULEID
ACTIONLIST
Example Usage: SecRuleRemoveByID 1 2 "9000-9010"
Processing Phase: Any
Scope: Any
Version: 2.0.0
Dependencies/Notes: This directive supports multiple parameters, where each parameter can either be a rule ID, or a range. Parameters that contain spaces must be delimited using double quotes.
SecRuleRemoveById 1 2 5 10-20 "400-556" 673
Description: Removes matching rules from the parent contexts.
Syntax: SecRuleRemoveByMsg REGEX
Example Usage: SecRuleRemoveByMsg "FAIL"
Processing Phase: Any
Scope: Any
Version: 2.0.0
Dependencies/Notes: This directive supports
multiple parameters. Each parameter is a regular expression that will be
applied to the message (specified using the msg
action).
Description: This directive creates a special
rule that executes a Lua script to decide whether to match or not. The
main difference from SecRule
is that there are no
targets nor operators. The script can fetch any variable from the
ModSecurity context and use any (Lua) operator to test them. The second
optional parameter is the list of actions whose meaning is identical to
that of SecRule
.
Syntax: SecRuleScript
/path/to/script.lua [ACTIONS]
Example Usage: SecRuleScript "/path/to/file.lua"
"block"
Processing Phase: Any
Scope: Any
Version: 2.5.0
Dependencies/Notes: None
All Lua scripts are compiled at configuration time and cached in memory. To reload scripts you must reload the entire ModSecurity configuration by restarting Apache.
Example script:
-- Your script must define the main entry -- point, as below. function main() -- Log something at level 1. Normally you shouldn't be -- logging anything, especially not at level 1, but this is -- just to show you can. Useful for debugging. m.log(1, "Hello world!"); -- Retrieve one variable. local var1 = m.getvar("REMOTE_ADDR"); -- Retrieve one variable, applying one transformation function. -- The second parameter is a string. local var2 = m.getvar("ARGS", "lowercase"); -- Retrieve one variable, applying several transformation functions. -- The second parameter is now a list. You should note that m.getvar() -- requires the use of comma to separate collection names from -- variable names. This is because only one variable is returned. local var3 = m.getvar("ARGS.p", { "lowercase", "compressWhitespace" } ); -- If you want this rule to match return a string -- containing the error message. The message must contain the name -- of the variable where the problem is located. -- return "Variable ARGS:p looks suspicious!" -- Otherwise, simply return nil. return nil; end
In this first example we were only retrieving one variable at the time. In this case the name of the variable is known to you. In many cases, however, you will want to examine variables whose names you won't know in advance, for example script parameters.
Example showing use of m.getvars()
to retrieve
many variables at once:
function main() -- Retrieve script parameters. local d = m.getvars("ARGS", { "lowercase", "htmlEntityDecode" } ); -- Loop through the paramters. for i = 1, #d do -- Examine parameter value. if (string.find(d[i].value, "<script")) then -- Always specify the name of the variable where the -- problem is located in the error message. return ("Suspected XSS in variable " .. d[i].name .. "."); end end -- Nothing wrong found. return nil; end
Go to http://www.lua.org/ to find more about the Lua programming language. The reference manual too is available online, at http://www.lua.org/manual/5.1/.
Lua support is marked as experimental as the way the progamming interface may continue to evolve while we are working for the best implementation style. Any user input into the programming interface is appreciated.
Description: Updates the action list of the specified rule.
Syntax: SecRuleRemoveById RULEID ACTIONLIST
Example Usage: SecRuleUpdateActionById 12345
deny,status:403
Processing Phase: Any
Scope: Any
Version: 2.5.0
Dependencies/Notes: This directive merges the specified action list with the rule's action list. There are two limitations. The rule ID cannot be changed, nor can the phase. Further note that actions that may be specified multiple times are appended to the original.
SecAction \ "t:lowercase,phase:2,id:12345,pass,msg:'The Message',log,auditlog" SecRuleUpdateActionById 12345 "t:compressWhitespace,deny,status:403,msg:'A new message'
The example above will cause the rule to be executed as if it was specified as follows:
SecAction \ "t:lowercase,phase:2,id:12345,log,auditlog,t:compressWhitespace,deny,status:403,msg:'A new message'"
Description: Instructs ModSecurity to change the data presented in the "Server:" response header token.
Syntax: SecServerSignature "WEB SERVER
SOFTWARE"
Example Usage: SecServerSignature
"Netscape-Enterprise/6.0"
Processing Phase: N/A
Scope: Main
Version: 2.0.0
Dependencies/Notes: In order for this directive to work, you must set the Apache ServerTokens directive to Full. ModSecurity will overwrite the server signature data held in this memory space with the data set in this directive. If ServerTokens is not set to Full, then the memory space is most likely not large enough to hold the new data we are looking to insert.
Description: Configures the directory where temporary files will be created.
Syntax: SecTmpDir
/path/to/dir
Example Usage: SecTmpDir /tmp
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: Needs to be writable by the Apache user process. This is the directory location where Apache will swap data to disk if it runs out of memory (more data than what was specified in the SecRequestBodyInMemoryLimit directive) during inspection.
Description: Configures the directory where intercepted files will be stored.
Syntax: SecUploadDir
/path/to/dir
Example Usage: SecUploadDir /tmp
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: This directory must be on
the same filesystem as the temporary directory defined with SecTmpDir
. This directive is used with
SecUploadKeepFiles
.
Description: Configures the mode (permissions) of any uploaded files using an octal mode (as used in chmod).
Syntax: SecUploadFileMode octal_mode|"default"
Example Usage: SecUploadFileMode 0640
Processing Phase: N/A
Scope: Any
Version: 2.1.6
Dependencies/Notes: This feature is not available on operating systems not supporting octal file modes. The default mode (0600) only grants read/write access to the account writing the file. If access from another account is needed (using clamd is a good example), then this directive may be required. However, use this directive with caution to avoid exposing potentially sensitive data to unauthorized users. Using the value "default" will revert back to the default setting.
The process umask may still limit the mode if it is being more restrictive than the mode set using this directive.
Description: Configures whether or not the intercepted files will be kept after transaction is processed.
Syntax: SecUploadKeepFiles On|Off|RelevantOnly
Example Usage: SecUploadKeepFiles On
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: This directive requires
the storage directory to be defined (using SecUploadDir
).
Possible values are:
On
- Keep uploaded
files.
Off
- Do not keep uploaded
files.
RelevantOnly
- This will
keep only those files that belong to requests that are deemed
relevant.
Description: Creates a partition on the server that belongs to one web application.
Syntax: SecWebAppId
"NAME"
Example Usage: SecWebAppId "WebApp1"
Processing Phase: N/A
Scope: Any
Version: 2.0.0
Dependencies/Notes: Partitions are used to
avoid collisions between session IDs and user IDs. This directive must
be used if there are multiple applications deployed on the same server.
If it isn't used, a collision between session IDs might occur. The
default value is default
.
Example:
<VirtualHost *:80> ServerName app1.com ServerAlias www.app1.com SecWebAppId "App1" SecRule REQUEST_COOKIES:PHPSESSID !^$ chain,nolog,pass SecAction setsid:%{REQUEST_COOKIES.PHPSESSID} ... </VirtualHost> <VirtualHost *:80> ServerName app2.com ServerAlias www.app2.com SecWebAppId "App2" SecRule REQUEST_COOKIES:PHPSESSID !^$ chain,nolog,pass SecAction setsid:%{REQUEST_COOKIES.PHPSESSID} ... </VirtualHost>
In the two examples configurations shown, SecWebAppId is being used in conjunction with the Apache VirtualHost directives. What this achieves is to create more unique collection names when being hosted on one server. Normally, when setsid is used, ModSecurity will create a collection with the name "SESSION" and it will hold the value specified. With using SecWebAppId as shown in the examples, however, the name of the collection would become "App1_SESSION" and "App2_SESSION".
SecWebAppId is relevant in two cases:
You are logging transactions/alerts to the ModSecurity Console and you want to use the web application ID to search only the transactions belonging to that application.
You are using the data persistence facility (collections SESSION and USER) and you need to avoid collisions between sessions and users belonging to different applications.
ModSecurity 2.x allows rules to be placed in one of the following five phases:
Request headers (REQUEST_HEADERS
)
Request body (REQUEST_BODY
)
Response headers (RESPONSE_HEADERS
)
Response body (RESPONSE_BODY
)
Logging (LOGGING
)
Below is a diagram of the standard Apache Request Cycle. In the diagram, the 5 ModSecurity processing phases are shown.
In order to select the phase a rule executes during, use the phase
action either directly in the rule or in using the
SecDefaultAction
directive:
SecDefaultAction "log,pass,phase:2" SecRule REQUEST_HEADERS:Host "!^$" "deny,phase:1"
Keep in mind that rules are executed according to phases, so even
if two rules are adjacent in a configuration file, but are set to
execute in different phases, they would not happen one after the other.
The order of rules in the configuration file is important only within
the rules of each phase. This is especially important when using the
skip
and skipAfter
actions.
The LOGGING
phase is special. It is executed at
the end of each transaction no matter what happened in the previous
phases. This means it will be processed even if the request was
intercepted or the allow
action was used to pass the
transaction through.
Rules in this phase are processed immediately after Apache completes reading the request headers (post-read-request phase). At this point the request body has not been read yet, meaning not all request arguments are available. Rules should be placed in this phase if you need to have them run early (before Apache does something with the request), to do something before the request body has been read, determine whether or not the request body should be buffered, or decide how you want the request body to be processed (e.g. whether to parse it as XML or not).
Note
Rules in this phase can not leverage Apache scope directives (Directory, Location, LocationMatch, etc...) as the post-read-request hook does not have this information yet. The exception here is the VirtualHost directive. If you want to use ModSecurity rules inside Apache locations, then they should run in Phase 2. Refer to the Apache Request Cycle/ModSecurity Processing Phases diagram.
This is the general-purpose input analysis phase. Most of the application-oriented rules should go here. In this phase you are guaranteed to have received the request arguments (provided the request body has been read). ModSecurity supports three encoding types for the request body phase:
application/x-www-form-urlencoded
- used to
transfer form data
multipart/form-data
- used for file
transfers
text/xml
- used for passing XML data
Other encodings are not used by most web applications.
This phase takes place just before response headers are sent back to the client. Run here if you want to observe the response before that happens, and if you want to use the response headers to determine if you want to buffer the response body. Note that some response status codes (such as 404) are handled earlier in the request cycle by Apache and my not be able to be triggered as expected. Additionally, there are some response headers that are added by Apache at a later hook (such as Date, Server and Connection) that we would not be able to trigger on or sanitize. This should work appropriately in a proxy setup or within phase:5 (logging).
This is the general-purpose output analysis phase. At this point you can run rules against the response body (provided it was buffered, of course). This is the phase where you would want to inspect the outbound HTML for information disclosure, error messages or failed authentication text.
This phase is run just before logging takes place. The rules placed into this phase can only affect how the logging is performed. This phase can be used to inspect the error messages logged by Apache. You cannot deny/block connections in this phase as it is too late. This phase also allows for inspection of other response headers that weren't available during phase:3 or phase:4. Note that you must be careful not to inherit a disruptive action into a rule in this phase as this is a configuration error in ModSecurity 2.5.0 and later versions.
The following variables are supported in ModSecurity 2.x:
ARGS
is a collection and can be used on its own
(means all arguments including the POST Payload), with a static
parameter (matches arguments with that name), or with a regular
expression (matches all arguments with name that matches the regular
expression). To look at only the query string or body arguments, see the
ARGS_GET
and ARGS_POST
collections.
Some variables are actually collections, which are expanded into more variables at runtime. The following example will examine all request arguments:
SecRule ARGS dirtySometimes, however, you will want to look only at parts of a collection. This can be achieved with the help of the selection operator(colon). The following example will only look at the arguments named
p
(do note that, in
general, requests can contain multiple arguments with the same name):
SecRule ARGS:p dirtyIt is also possible to specify exclusions. The following will examine all request arguments for the word dirty, except the ones named
z
(again, there can be
zero or more arguments named z
):
SecRule ARGS|!ARGS:z dirtyThere is a special operator that allows you to count how many variables there are in a collection. The following rule will trigger if there is more than zero arguments in the request (ignore the second parameter for the time being):
SecRule &ARGS !^0$And sometimes you need to look at an array of parameters, each with a slightly different name. In this case you can specify a regular expression in the selection operator itself. The following rule will look into all arguments whose names begin with
id_
: SecRule ARGS:/^id_/ dirty
Using ARGS:p
will not result in any
invocations against the operator if argument p does not exist.
In ModSecurity 1.X, the ARGS
variable stood
for QUERY_STRING
+ POST_PAYLOAD
,
whereas now it expands to individual variables.
This variable allows you to set more targeted evaluations on the total size of the Arguments as compared with normal Apache LimitRequest directives. For example, you could create a rule to ensure that the total size of the argument data is below a certain threshold (to help prevent buffer overflow issues). Example: Block request if the size of the arguments is above 25 characters.
SecRule REQUEST_FILENAME "^/cgi-bin/login\.php" \
"chain,log,deny,phase:2,t:none,t:lowercase,t:normalisePath"
SecRule ARGS_COMBINED_SIZE "@gt 25"
Is a collection of the argument names. You can search for specific argument names that you want to block. In a positive policy scenario, you can also whitelist (using an inverted rule with the ! character) only authorized argument names. Example: This example rule will only allow 2 argument names - p and a. If any other argument names are injected, it will be blocked.
SecRule REQUEST_FILENAME "/index.php" \
"chain,log,deny,status:403,phase:2,t:none,t:lowercase,t:normalisePath"
SecRule ARGS_NAMES "!^(p|a)$" "t:none,t:lowercase"
ARGS_GET_NAMES
is similar to
ARGS_NAMES
, but only contains argument names from the
query string.
ARGS_POST_NAMES
is similar to
ARGS_NAMES
, but only contains argument names from the
POST body.
This variable holds the authentication method used to validate a user. Example:
SecRule AUTH_TYPE "basic" log,deny,status:403,phase:1,t:lowercase
Note
This data will not be available in a proxy-mode deployment as the
authentication is not local. In a proxy-mode deployment, you would need
to inspect the REQUEST_HEADERS:Authorization
header.
Collection, requires a single parameter (after colon). The
ENV
variable is set with setenv and does not give
access to the CGI environment variables. Example:
SecRule REQUEST_FILENAME "printenv" pass,setenv:tag=suspicious SecRule ENV:tag "suspicious"
Collection. Contains a collection of original file names (as they were called on the remote user's file system). Note: only available if files were extracted from the request body. Example:
SecRule FILES "\.conf$" log,deny,status:403,phase:2
Single value. Total size of the uploaded files. Note: only available if files were extracted from the request body. Example:
SecRule FILES_COMBINED_SIZE "@gt 1000" log,deny,status:403,phase:2
Collection w/o parameter. Contains a list of form fields that were used for file upload. Note: only available if files were extracted from the request body. Example:
SecRule FILES_NAMES "^upfile$" log,deny,status:403,phase:2
Collection. Contains a list of file sizes. Useful for implementing a size limitation on individual uploaded files. Note: only available if files were extracted from the request body. Example:
SecRule FILES_SIZES "@gt 100" log,deny,status:403,phase:2
Collection. Contains a collection of temporary files' names on the
disk. Useful when used together with @inspectFile.
Note: only available if files
were extracted from the request body. Example:
SecRule FILES_TMPNAMES "@inspectFile /path/to/inspect_script.pl"
GEO
is a collection populated by the results of
the last @geoLookup
operator. The
collection can be used to match geographical fields looked from an IP
address or hostname.
Available since ModSecurity 2.5.0.
Fields:
COUNTRY_CODE: Two character country code. EX: US, GB, etc.
COUNTRY_CODE3: Up to three character country code.
COUNTRY_NAME: The full country name.
COUNTRY_CONTINENT: The two character continent that the country is located. EX: EU
REGION: The two character region. For US, this is state. For Canada, providence, etc.
CITY: The city name if supported by the database.
POSTAL_CODE: The postal code if supported by the database.
LATITUDE: The latitude if supported by the database.
LONGITUDE: The longitude if supported by the database.
DMA_CODE: The metropolitan area code if supported by the database. (US only)
AREA_CODE: The phone system area code. (US only)
Example:
SecGeoLookupDb /usr/local/geo/data/GeoLiteCity.dat
...
SecRule REMOTE_ADDR "@geoLookup" "chain,drop,msg:'Non-GB IP address'"
SecRule GEO:COUNTRY_CODE "!@streq GB"
This variable holds the highest severity of any rules that have
matched so far. Severities are numeric values and thus can be used with
comparison operators such as @lt
,
etc.
Higher severities have a lower numeric value.
A value of 255 indicates no severity has been set.
SecRule HIGHEST_SEVERITY "@le 2" "phase:2,deny,status:500,msg:'severity %{HIGHEST_SEVERITY}'"
This variable holds the value of the variable that was matched
against. It is similar to the TX:0, except it can be used for all
operators and does not require that the capture
action be specified.
SecRule ARGS pattern chain,deny
...
SecRule MATCHED_VAR "further scrutiny"
This variable holds the full name of the variable that was matched against.
SecRule ARGS pattern setvar:tx.mymatch=%{MATCHED_VAR_NAME}
...
SecRule TX:MYMATCH "@eq ARGS:param" deny
This variable holds the ModSecurity build number. This variable is intended to be used to check the build number prior to using a feature that is available only in a certain build. Example:
SecRule MODSEC_BUILD "!@ge 02050102" skipAfter:12345
SecRule ARGS "@pm some key words" id:12345,deny,status:500
This flag variable will be set to 1
whenever a
multi-part request uses mixed line terminators. The
multipart/form-data
RFC requires
CRLF
sequence to be used to terminate lines. Since
some client implementations use only LF
to terminate
lines you might want to allow them to proceed under certain
circumstances (if you want to do this you will need to stop using
MULTIPART_STRICT_ERROR
and check each multi-part flag
variable individually, avoiding MULTIPART_LF_LINE
).
However, mixing CRLF
and LF
line
terminators is dangerous as it can allow for evasion. Therefore, in such
cases, you will have to add a check for
MULTIPART_CRLF_LF_LINES
.
MULTIPART_STRICT_ERROR
will be set to
1
when any of the following variables is also set to
1
: REQBODY_PROCESSOR_ERROR
,
MULTIPART_BOUNDARY_QUOTED
,
MULTIPART_BOUNDARY_WHITESPACE
,
MULTIPART_DATA_BEFORE
,
MULTIPART_DATA_AFTER
,
MULTIPART_HEADER_FOLDING
,
MULTIPART_LF_LINE
,
MULTIPART_SEMICOLON_MISSING
MULTIPART_INVALID_QUOTING
. Each of these variables
covers one unusual (although sometimes legal) aspect of the request body
in multipart/form-data format
. Your policies should
always contain a rule to check either this variable
(easier) or one or more individual variables (if you know exactly what
you want to accomplish). Depending on the rate of false positives and
your default policy you should decide whether to block or just warn when
the rule is triggered.
The best way to use this variable is as in the example below:
SecRule MULTIPART_STRICT_ERROR "!@eq 0" \ "phase:2,t:none,log,deny,msg:'Multipart request body \ failed strict validation: \ PE %{REQBODY_PROCESSOR_ERROR}, \ BQ %{MULTIPART_BOUNDARY_QUOTED}, \ BW %{MULTIPART_BOUNDARY_WHITESPACE}, \ DB %{MULTIPART_DATA_BEFORE}, \ DA %{MULTIPART_DATA_AFTER}, \ HF %{MULTIPART_HEADER_FOLDING}, \ LF %{MULTIPART_LF_LINE}, \ SM %{MULTIPART_SEMICOLON_MISSING}, \ IQ %{MULTIPART_INVALID_QUOTING}'"
The multipart/form-data
parser was upgraded in
ModSecurity v2.1.3 to actively look for signs of evasion. Many variables
(as listed above) were added to expose various facts discovered during
the parsing process. The MULTIPART_STRICT_ERROR
variable is handy to check on all abnormalities at once. The individual
variables allow detection to be fine-tuned according to your
circumstances in order to reduce the number of false positives. Detailed
analysis of various evasion techniques covered will be released as a
separated document at a later date.
Set to 1
when, during the parsing phase of a
multipart/request-body
, ModSecurity encounters what
feels like a boundary but it is not. Such an event may occur when
evasion of ModSecurity is attempted.
The best way to use this variable is as in the example below:
SecRule MULTIPART_UNMATCHED_BOUNDARY "!@eq 0" \ "phase:2,t:none,log,deny,msg:'Multipart parser detected a possible unmatched boundary.'"
Change the rule from blocking to logging-only if many false positives are encountered.
Besides passing query information to a script/handler, you can also pass additional data, known as extra path information, as part of the URL. Example:
SecRule PATH_INFO "^/(bin|etc|sbin|opt|usr)"
This variable holds form data passed to the script/handler by appending data after a question mark. Warning: Not URL-decoded. Example:
SecRule QUERY_STRING "attack"
This variable holds the IP address of the remote client. Example:
SecRule REMOTE_ADDR "^192\.168\.1\.101$"
If HostnameLookUps are set to On, then this variable will hold the DNS resolved remote host name. If it is set to Off, then it will hold the remote IP address. Possible uses for this variable would be to deny known bad client hosts or network blocks, or conversely, to allow in authorized hosts. Example:
SecRule REMOTE_HOST "\.evil\.network\org$"
This variable holds information on the source port that the client
used when initiating the connection to our web server. Example: in this
example, we are evaluating to see if the REMOTE_PORT
is less than 1024, which would indicate that the user is a privileged
user (root).
SecRule REMOTE_PORT "@lt 1024" phase:1,log,pass,setenv:remote_port=privileged
This variable holds the username of the authenticated user. If there are no password (basic|digest) access controls in place, then this variable will be empty. Example:
SecRule REMOTE_USER "admin"
Note
This data will not be available in a proxy-mode deployment as the authentication is not local.
Built-in processors are URLENCODED
,
MULTIPART
, and XML
.
Example:
SecRule REQBODY_PROCESSOR "^XML$ chain
SecRule XML "@validateDTD /opt/apache-frontend/conf/xml.dtd"
Possible values are 0 (no error) or 1 (error). This variable will
be set by request body processors (typically the
multipart/request-data
parser or the XML parser)
when they fail to properly parse a request payload.
Example:
SecRule REQBODY_PROCESSOR_ERROR "@eq 1" deny,phase:2
Your policies must have a rule to check REQBODY_PROCESSOR_ERROR at the beginning of phase 2. Failure to do so will leave the door open for impedance mismatch attacks. It is possible, for example, that a payload that cannot be parsed by ModSecurity can be successfully parsed by more tolerant parser operating in the application. If your policy dictates blocking then you should reject the request if error is detected. When operating in detection-only mode your rule should alert with high severity when request body processing fails.
Empty, or contains the error message from the processor. Example:
SecRule REQBODY_PROCESSOR_ERROR_MSG "failed to parse" t:lowercase
This variable holds just the filename part of
REQUEST_FILENAME
(e.g. index.php).
Example:
SecRule REQUEST_BASENAME "^login\.php$" phase:2,t:none,t:lowercase
Please note that anti-evasion transformations are not applied to
this variable by default. REQUEST_BASENAME
will
recognise both /
and \
as path
separators.
This variable holds the data in the request body (including
POST_PAYLOAD
data). REQUEST_BODY
should be used if the original order of the arguments is important
(ARGS
should be used in all other cases).
Example:
SecRule REQUEST_BODY "^username=\w{25,}\&password=\w{25,}\&Submit\=login$"
This variable is only available if the
URLENCODED
request body processor parsed a request
body. This will occur by default when an
application/x-www-form-urlencoded
is detected, or
the URLENCODED
request body parser is forced. As of
2.5.7 it is possible to force the presence of the
REQUEST_BODY
variable, but only when there is no
request body processor defined, using the
ctl:forceRequestBodyVariable
option in the
REQUEST_HEADERS
phase.
This variable is a collection of all of the cookie data. Example: the following example is using the Ampersand special operator to count how many variables are in the collection. In this rule, it would trigger if the request does not include any Cookie headers.
SecRule &REQUEST_COOKIES "@eq 0"
This variable is a collection of the cookie names in the request headers. Example: the following rule will trigger if the JSESSIONID cookie is not present.
SecRule &REQUEST_COOKIES_NAMES:JSESSIONID "@eq 0"
This variable holds the relative REQUEST_URI
minus the QUERY_STRING
part (e.g. /index.php).
Example:
SecRule REQUEST_FILENAME "^/cgi-bin/login\.php$" phase:2,t:none,t:normalisePath
Please note that anti-evasion transformations are not used on
REQUEST_FILENAME
by default.
This variable can be used as either a collection of all of the
request headers or can be used to specify individual headers (by using
REQUEST_HEADERS:Header-Name). Example: the first
example uses REQUEST_HEADERS
as a collection and is
applying the validateUrlEncoding
operator against all
headers.
SecRule REQUEST_HEADERS "@validateUrlEncoding"
Example: the second example is targeting only the
Host
header.
SecRule REQUEST_HEADERS:Host "^[\d\.]+$" \
"deny,log,status:400,msg:'Host header is a numeric IP address'"
This variable is a collection of the names of all of the request headers. Example:
SecRule REQUEST_HEADERS_NAMES "^x-forwarded-for" \
"log,deny,status:403,t:lowercase,msg:'Proxy Server Used'"
This variable holds the complete request line sent to the server (including the REQUEST_METHOD and HTTP version data). Example: this example rule will trigger if the request method is something other than GET, HEAD, POST or if the HTTP is something other than HTTP/0.9, 1.0 or 1.1.
SecRule REQUEST_LINE "!(^((?:(?:pos|ge)t|head))|http/(0\.9|1\.0|1\.1)$)" t:none,t:lowercase
This variable holds the request method used by the client.
The following example will trigger if the request method is either
CONNECT
or TRACE.
SecRule REQUEST_METHOD "^((?:connect|trace))$" t:none,t:lowercase
This variable holds the request protocol version information. Example:
SecRule REQUEST_PROTOCOL "!^http/(0\.9|1\.0|1\.1)$" t:none,t:lowercase
This variable holds the full URL including the
QUERY_STRING
data (e.g. /index.php?p=X), however it
will never contain a domain name, even if it was provided on the request
line. It also does not include either the
REQUEST_METHOD
or the HTTP version info.
Example:
SecRule REQUEST_URI "attack" phase:1,t:none,t:urlDecode,t:lowercase,t:normalisePath
Please note that anti-evasion transformations are not used on
REQUEST_URI
by default.
Same as REQUEST_URI
but will contain the domain
name if it was provided on the request line (e.g.
http://www.example.com/index.php?p=X).
Example:
SecRule REQUEST_URI_RAW "http:/" phase:1,t:none,t:urlDecode,t:lowercase,t:normalisePath
Please note that anti-evasion transformations are not used on
REQUEST_URI_RAW
by default.
This variable holds the data for the response payload.
Example:
SecRule RESPONSE_BODY "ODBC Error Code"
Response body length in bytes. Can be available starting with
phase 3 but it does not have to be (as the length of response body is
not always known in advance.) If the size is not known this variable
will contain a zero. If RESPONSE_CONTENT_LENGTH
contains a zero in phase 5 that means the actual size of the response
body was 0.
The value of this variable can change between phases if the body
is modified. For example, in embedded mode
mod_deflate
can compress the response body between
phases 4 and 5.
This variable is similar to the REQUEST_HEADERS variable and can be used in the same manner. Example:
SecRule RESPONSE_HEADERS:X-Cache "MISS"
Note
This variable may not have access to some headers when running in embedded-mode. Headers such as Server, Date, Connection and Content-Type are added during a later Apache hook just prior to sending the data to the client. This data should be available, however, either during ModSecurity phase:5 (logging) or when running in proxy-mode.
This variable is a collection of the response header names. Example:
SecRule RESPONSE_HEADERS_NAMES "Set-Cookie"
Note
Same limitations as RESPONSE_HEADERS with regards to access to some headers in embedded-mode.
This variable holds the HTTP response protocol information. Example:
SecRule RESPONSE_PROTOCOL "^HTTP\/0\.9"
This variable holds the HTTP response status code as generated by Apache. Example:
SecRule RESPONSE_STATUS "^[45]"
Note
This directive may not work as expected in embedded-mode as Apache handles many of the stock response codes (404, 401, etc...) earlier in Phase 2. This variable should work as expected in a proxy-mode deployment.
This variable provides access to the id
, rev
,
severity
, logdata
, and msg
fields of the rule that triggered the
action. Only available for expansion in action strings (e.g.setvar:tx.varname=%{rule.id}
). Example:
SecRule &REQUEST_HEADERS:Host "@eq 0" "log,deny,setvar:tx.varname=%{rule.id}"
This variable holds just the local filename part of SCRIPT_FILENAME. Example:
SecRule SCRIPT_BASENAME "^login\.php$"
Note
This variable is not available in proxy mode.
This variable holds the full path on the server to the requested script. (e.g. SCRIPT_NAME plus the server path). Example:
SecRule SCRIPT_FILENAME "^/usr/local/apache/cgi-bin/login\.php$"
Note
This variable is not available in proxy mode.
This variable holds the group id (numerical value) of the group owner of the script. Example:
SecRule SCRIPT_GID "!^46$"
Note
This variable is not available in proxy mode.
This variable holds the group name of the group owner of the script. Example:
SecRule SCRIPT_GROUPNAME "!^apache$"
Note
This variable is not available in proxy mode.
This variable holds the script's permissions mode data (numerical - 1=execute, 2=write, 4=read and 7=read/write/execute). Example: will trigger if the script has the WRITE permissions set.
SecRule SCRIPT_MODE "^(2|3|6|7)$"
Note
This variable is not available in proxy mode.
This variable holds the user id (numerical value) of the owner of the script. Example: the example rule below will trigger if the UID is not 46 (the Apache user).
SecRule SCRIPT_UID "!^46$"
Note
This variable is not available in proxy mode.
This variable holds the username of the owner of the script. Example:
SecRule SCRIPT_USERNAME "!^apache$"
Note
This variable is not available in proxy mode.
This variable contains the IP address of the server. Example:
SecRule SERVER_ADDR "^192\.168\.1\.100$"
This variable contains the server's hostname or IP address. Example:
SecRule SERVER_NAME "hostname\.com$"
Note
This data is taken from the Host header submitted in the client request.
This variable contains the local port that the web server is listening on. Example:
SecRule SERVER_PORT "^80$"
This variable is a collection, available only after setsid
is executed. Example: the following
example shows how to initialize a SESSION collection with setsid, how to
use setvar to increase the session.score values, how to set the
session.blocked variable and finally how to deny the connection based on
the session:blocked value.
SecRule REQUEST_COOKIES:PHPSESSID !^$ chain,nolog,pass SecAction setsid:%{REQUEST_COOKIES.PHPSESSID} SecRule REQUEST_URI "^/cgi-bin/finger$" \ "phase:2,t:none,t:lowercase,t:normalisePath,pass,log,setvar:session.score=+10" SecRule SESSION:SCORE "@gt 50" "pass,log,setvar:session.blocked=1" SecRule SESSION:BLOCKED "@eq 1" "log,deny,status:403"
This variable is the value set with setsid
. Example:
SecRule SESSIONID !^$ chain,nolog,pass
SecRule REQUEST_COOKIES:PHPSESSID !^$
SecAction setsid:%{REQUEST_COOKIES.PHPSESSID}
This variable holds a formatted string representing the time (hour:minute:second). Example:
SecRule TIME "^(([1](8|9))|([2](0|1|2|3))):\d{2}:\d{2}$"
This variable holds the current date (1-31). Example: this rule would trigger anytime between the 10th and 20th days of the month.
SecRule TIME_DAY "^(([1](0|1|2|3|4|5|6|7|8|9))|20)$"
This variable holds the current hour (0-23). Example: this rule would trigger during "off hours".
SecRule TIME_HOUR "^(0|1|2|3|4|5|6|[1](8|9)|[2](0|1|2|3))$"
This variable holds the current minute (0-59). Example: this rule would trigger during the last half hour of every hour.
SecRule TIME_MIN "^(3|4|5)"
This variable holds the current month (0-11). Example: this rule would match if the month was either November (10) or December (11).
SecRule TIME_MON "^1"
This variable holds the current weekday (0-6). Example: this rule would trigger only on week-ends (Saturday and Sunday).
SecRule TIME_WDAY "^(0|6)$"
Transaction Collection. This is used to store pieces of data, create a transaction anomaly score, and so on. Transaction variables are set for 1 request/response cycle. The scoring and evaluation will not last past the current request/response process. Example: In this example, we are using setvar to increase the tx.score value by 5 points. We then have a follow-up run that will evaluate the transactional score this request and then it will decided whether or not to allow/deny the request through.
The following is a list of reserved names in the TX collection:
TX:0
- The matching value
when using the @rx
or @pm
operator with the capture
action.
TX:1-TX:9
- The captured
subexpression value when using the @rx
operator with capturing parens and the
capture
action.
SecRule WEBSERVER_ERROR_LOG "does not exist" "phase:5,pass,setvar:tx.score=+5" SecRule TX:SCORE "@gt 20" deny,log
This variable is the value set with setuid
. Example:
SecAction setuid:%{REMOTE_USER},nolog
SecRule USERID "Admin"
This variable is the value set with SecWebAppId
. Example:
SecWebAppId "WebApp1"
SecRule WEBAPPID "WebApp1" "chain,log,deny,status:403"
SecRule REQUEST_HEADERS:Transfer-Encoding "!^$"
Contains zero or more error messages produced by the web server. Access to this variable is in phase:5 (logging). Example:
SecRule WEBSERVER_ERROR_LOG "File does not exist" "phase:5,setvar:tx.score=+5"
Can be used standalone (as a target for
validateDTD
and validateSchema
) or
with an XPath expression parameter (which makes it a valid target for
any function that accepts plain text). Example using XPath:
SecDefaultAction log,deny,status:403,phase:2 SecRule REQUEST_HEADERS:Content-Type ^text/xml$ \ phase:1,t:lowercase,nolog,pass,ctl:requestBodyProcessor=XML SecRule REQBODY_PROCESSOR "!^XML$" skipAfter:12345 SecRule XML:/employees/employee/name/text() Fred SecRule XML:/xq:employees/employee/name/text() Fred \ id:12345,xmlns:xq=http://www.example.com/employees
The first XPath expression does not use namespaces. It would match against payload such as this one:
<employees> <employee> <name>Fred Jones</name> <address location="home"> <street>900 Aurora Ave.</street> <city>Seattle</city> <state>WA</state> <zip>98115</zip> </address> <address location="work"> <street>2011 152nd Avenue NE</street> <city>Redmond</city> <state>WA</state> <zip>98052</zip> </address> <phone location="work">(425)555-5665</phone> <phone location="home">(206)555-5555</phone> <phone location="mobile">(206)555-4321</phone> </employee> </employees>
The second XPath expression does use namespaces. It would match the following payload:
<xq:employees xmlns:xq="http://www.example.com/employees"> <employee> <name>Fred Jones</name> <address location="home"> <street>900 Aurora Ave.</street> <city>Seattle</city> <state>WA</state> <zip>98115</zip> </address> <address location="work"> <street>2011 152nd Avenue NE</street> <city>Redmond</city> <state>WA</state> <zip>98052</zip> </address> <phone location="work">(425)555-5665</phone> <phone location="home">(206)555-5555</phone> <phone location="mobile">(206)555-4321</phone> </employee> </xq:employees>
Note the different namespace used in the second example.
To learn more about XPath we suggest the following resources:
When ModSecurity receives request or response information, it makes a copy of this data and places it into memory. It is on this data in memory that transformation functions are applied. The raw request/response data is never altered. Transformation functions are used to transform a variable before testing it in a rule.
Note
There are no default transformation functions as there were in previous versions of ModSecurity.
The following rule will ensure that an attacker does not use mixed case in order to evade the ModSecurity rule:
SecRule ARGS:p "xp_cmdshell" "t:lowercase"
multiple transformation actions can be used in the same rule, for example
the following rule also ensures that an attacker does not use URL encoding
(%xx encoding) for evasion. Note the order of the transformation
functions, which ensures that a URL encoded letter is first decoded and
than translated to lower case.SecRule ARGS:p "xp_cmdshell" "t:urlDecode,t:lowercase"
One can use the SecDefaultAction command to ensure the translation occurs for every rule until the next. Note that transformation actions are additive, so if a rule explicitly list actions, the translation actions set by SecDefaultAction are still performed.
SecDefaultAction t:urlDecode,t:lowercase
The following transformation functions are supported:
It converts whitespace characters (32, \f, \t, \n, \r, \v, 160) to spaces (ASCII 32) and then compresses multiple consecutive space characters into one.
Decodes CSS-encoded characters, as specified at http://www.w3.org/TR/REC-CSS2/syndata.html.
This function uses only up to two bytes in the decoding process, meaning
it is useful to uncover ASCII characters (that wouldn't normally be
encoded) encoded using CSS encoding, or to counter evasion which is a
combination of a backslash and non-hexadecimal characters (e.g.
ja\vascript
is equivalent to
javascript
).
This function decode ANSI C escape sequences: \a
, \b
,
\f
, \n
, \r
,
\t
, \v
, \\
,
\?
, \'
, \"
,
\xHH
(hexadecimal), \0OOO
(octal). Invalid encodings are left in
the output.
This function decodes HTML entities present in input. The following variants are supported:
&#xHH
and &#xHH;
(where H is any hexadecimal
number)
&#DDD
and &#DDD;
(where D is any decimal
number)
"
and "
 
and
<
and <
>
and >
This function will convert any entity into a single byte only, possibly resulting in a loss of information. It is thus useful to uncover bytes that would otherwise not need to be encoded, but it cannot do anything with the characters from the range above 255.
Decodes JavaScript escape sequences. If a
\uHHHH
code is in the range of
FF01
-FF5E
(the full width ASCII
codes), then the higher byte is used to detect and adjust the lower
byte. Otherwise, only the lower byte will be used and the higher byte
zeroed.
This function calculates an MD5 hash from input. Note that the
computed hash is in a raw binary form and may need encoded into text to
be usable (for example: t:md5,t:hexEncode
).
Not an actual transformation function, but an instruction to ModSecurity to remove all transformation functions associated with the current rule.
This function will remove multiple slashes, self-references and directory back-references (except when they are at the beginning of the input).
Same as normalisePath
, but will first convert
backslash characters to forward slashes.
This function calculates even parity of 7-bit data replacing the 8th bit of each target byte with the calculated parity bit.
This function calculates odd parity of 7-bit data replacing the 8th bit of each target byte with the calculated parity bit.
This function calculates zero parity of 7-bit data replacing the 8th bit of each target byte with a zero parity bit which allows inspection of even/odd parity 7bit data as ASCII7 data.
This function replaces each occurrence of a C-style comments
(/* ... */
) with a single space
(multiple consecutive occurrences of a space will not be compressed).
Unterminated comments will too be replaced with a space (ASCII 32).
However, a standalone termination of a comment (*/
) will not be acted upon.
This function is enabled by default. It replaces NULL bytes in input with spaces (ASCII 32).
This function decodes an URL-encoded input string. Invalid
encodings (i.e. the ones that use non-hexadecimal characters, or the
ones that are at the end of string and have one or two characters
missing) will not be converted. If you want to detect invalid encodings
use the @validateUrlEncoding
operator. The transformation function should not be used against
variables that have already been URL-decoded unless it is your intention
to perform URL decoding twice!
In addition to decoding %xx
like urlDecode, urlDecodeUni
also decodes %uXXXX
encoding. If the code is in the range
of FF01
-FF5E
(the full width ASCII
codes), then the higher byte is used to detect and adjust the lower
byte. Otherwise, only the lower byte will be used and the higher byte
zeroed.
This function calculates a SHA1 hash from input. Note that the
computed hash is in a raw binary form and may need encoded to be usable
(for example: t:sha1,t:hexEncode
).
Each action belongs to one of five groups:
Cause ModSecurity to do something. In many cases something means block transaction, but not in all. For example, the allow action is classified as a disruptive action, but it does the opposite of blocking. There can only be one disruptive action per rule (if there are multiple disruptive actions present, or inherited, only the last one will take effect), or rule chain (in a chain, a disruptive action can only appear in the first rule).
Do something, but that something does not and cannot affect the rule processing flow. Setting a variable, or changing its value is an example of a non-disruptive action. Non-disruptive action can appear in any rule, including each rule belonging to a chain.
These actions affect the rule flow (for example
skip
or skipAfter
).
Meta-data actions are used to provide more information about
rules. Examples include id
,
rev
, severity
and
msg
.
Not really actions, these are mere containers that hold data
used by other actions. For example, the status
action holds the status that will be used for blocking (if it takes
place).
Description: Stops rule processing on a successful match and allows the transaction to proceed.
Action Group: Disruptive
Example:
SecRule REMOTE_ADDR "^192\.168\.1\.100$" nolog,phase:1,allow
Prior to ModSecurity 2.5 the allow
action would
only affect the current phase. An allow
in phase 1
would skip processing the remaining rules in phase 1 but the rules from
phase 2 would execute. Starting with v2.5.0 allow
was
enhanced to allow for fine-grained control of what is done. The
following rules now apply:
If used one its own, like in the example above,
allow
will affect the entire transaction,
stopping processing of the current phase but also skipping over all
other phases apart from the logging phase. (The logging phase is
special; it is designed to always execute.)
If used with parameter "phase", allow
will
cause the engine to stop processing the current phase. Other phases
will continue as normal.
If used with parameter "request", allow
will cause the engine to stop processing the current phase. The next
phase to be processed will be phase
RESPONSE_HEADERS
.
Examples:
# Do not process request but process response. SecAction phase:1,allow:request # Do not process transaction (request and response). SecAction phase:1,allow
If you want to allow a response through, put a rule in phase
RESPONSE_HEADERS
and simply use
allow
on its own:
# Allow response through. SecAction phase:3,allow
Description: Appends text given as parameter
to the end of response body. For this action to work content injection
must be enabled by setting SecContentInjection
to
On
. Also make sure you check the content type of the
response before you make changes to it (e.g. you don't want to inject
stuff into images).
Action Group: Non-disruptive
Processing Phases: 3 and 4.
Example:
SecRule RESPONSE_CONTENT_TYPE "^text/html" "nolog,pass,append:'<hr>Footer'"
While macro expansion is allowed in the additional content, you are strongly cautioned against inserting user defined data fields.
Description: Marks the transaction for logging in the audit log.
Action Group: Non-disruptive
Example:
SecRule REMOTE_ADDR "^192\.168\.1\.100$" auditlog,phase:1,allow
Note
The auditlog action is now explicit if log is already specified.
Description: Performs the default disruptive action.
Action Group: Disruptive
It is intended to be used by ruleset writers to signify that the
rule was intended to block and leaves the "how" up to the administrator.
This action is currently a placeholder which will just be replaced by
the action from the last SecDefaultAction
in the same
context. Using the block
action with the
SecRuleUpdateActionById
directive allows a rule to be
reverted back to the previous SecDefaultAction
disruptive action.
In future versions of ModSecurity, more control and functionality will be added to define "how" to block.
Examples:
In the following example, the second rule will "deny" because of the SecDefaultAction disruptive action. The intent being that the administrator could easily change this to another disruptive action without editing the actual rules.
### Administrator defines "how" to block (deny,status:403)... SecDefaultAction phase:2,deny,status:403,log,auditlog ### Included from a rulest... # Intent is to warn for this User Agent SecRule REQUEST_HEADERS:User-Agent "perl" "phase:2,pass,msg:'Perl based user agent identified'" # Intent is to block for this User Agent, "how" described in SecDefaultAction SecRule REQUEST_HEADERS:User-Agent "nikto" "phase:2,block,msg:'Nikto Scanners Identified'"
In the following example, The rule is reverted back to the
pass
action defined in the SecDefaultAction directive
by using the SecRuleUpdateActionById
directive in
conjuction with the block
action. This allows an
administrator to override an action in a 3rd party rule without
modifying the rule itself.
### Administrator defines "how" to block (deny,status:403)... SecDefaultAction phase:2,pass,log,auditlog ### Included from a rulest... SecRule REQUEST_HEADERS:User-Agent "nikto" "id:1,phase:2,deny,msg:'Nikto Scanners Identified'" ### Added by the administrator SecRuleUpdateActionById 1 "block"
Description: When used together with the regular expression operator, capture action will create copies of regular expression captures and place them into the transaction variable collection. Up to ten captures will be copied on a successful pattern match, each with a name consisting of a digit from 0 to 9.
Action Group: Non-disruptive
Example:
SecRule REQUEST_BODY "^username=(\w{25,})" phase:2,capture,t:none,chain
SecRule TX:1 "(?:(?:a(dmin|nonymous)))"
Note
The 0 data captures the entire REGEX match and 1 captures the data in the first parens, etc...
Description: Chains the rule where the action is placed with the rule that immediately follows it. The result is called a rule chain. Chained rules allow for more complex rule matches where you want to use a number of different VARIABLES to create a better rule and to help prevent false positives.
Action Group: Flow
Example:
# Refuse to accept POST requests that do
# not specify request body length. Do note that
# this rule should be preceeded by a rule that verifies
# only valid request methods (e.g. GET, HEAD and POST) are used.
SecRule REQUEST_METHOD ^POST$ chain,t:none
SecRule REQUEST_HEADERS:Content-Length ^$ t:none
In programming language concepts, think of chained rules somewhat similar to AND conditional statements. The actions specified in the first portion of the chained rule will only be triggered if all of the variable checks return positive hits. If one aspect of the chained rule is negative, then the entire rule chain is negative. Also note that disruptive actions, execution phases, metadata actions (id, rev, msg), skip and skipAfter actions can only be specified on by the chain starter rule.
Description: The ctl action allows configuration options to be updated for the transaction.
Action Group: Non-disruptive
Example:
# Parse requests with Content-Type "text/xml" as XML
SecRule REQUEST_CONTENT_TYPE ^text/xml nolog,pass,ctl:requestBodyProcessor=XML
Note
The following configuration options are supported:
auditEngine
auditLogParts
debugLogLevel
ruleRemoveById
(single rule
ID, or a single rule ID range accepted as parameter)
requestBodyAccess
forceRequestBodyVariable
requestBodyLimit
requestBodyProcessor
responseBodyAccess
responseBodyLimit
ruleEngine
With the exception of
requestBodyProcessor
and
forceRequestBodyVariable
, each configuration option
corresponds to one configuration directive and the usage is
identical.
The requestBodyProcessor
option allows you to
configure the request body processor. By default ModSecurity will use
the URLENCODED
and MULTIPART
processors to process an application/x-www-form-urlencoded
and a
multipart/form-data
bodies,
respectively. A third processor, XML
, is also
supported, but it is never used implicitly. Instead you must tell
ModSecurity to use it by placing a few rules in the REQUEST_HEADERS
processing phase. After the
request body was processed as XML you will be able to use the
XML-related features to inspect it.
Request body processors will not interrupt a transaction if an
error occurs during parsing. Instead they will set variables REQBODY_PROCESSOR_ERROR
and REQBODY_PROCESSOR_ERROR_MSG
. These variables
should be inspected in the REQUEST_BODY
phase and an appropriate action
taken.
The forceRequestBodyVariable
option allows you
to configure the REQUEST_BODY
variable to be set when
there is no request body processor configured. This allows for
inspection of request bodies of unknown types.
Description: Stops rule processing and intercepts transaction.
Action Group: Disruptive
Example:
SecRule REQUEST_HEADERS:User-Agent "nikto" "log,deny,msg:'Nikto Scanners Identified'"
Description: Decrement counter based on its age.
Action Group: Non-Disruptive
Example: The following example will decrement the counter by 60 every 300 seconds.
SecAction deprecatevar:session.score=60/300
Note
Counter values are always positive, meaning the value will never go below zero.
Description: Immediately initiate a "connection close" action to tear down the TCP connection by sending a FIN packet.
Action Group: Disruptive
Example: The following example initiates an IP collection for tracking Basic Authentication attempts. If the client goes over the threshold of more than 25 attempts in 2 minutes, it will DROP subsequent connections.
SecAction phase:1,initcol:ip=%{REMOTE_ADDR},nolog
SecRule ARGS:login "!^$" \
nolog,phase:1,setvar:ip.auth_attempt=+1,deprecatevar:ip.auth_attempt=20/120
SecRule IP:AUTH_ATTEMPT "@gt 25" \
"log,drop,phase:1,msg:'Possible Brute Force Attack'"
Note
This action is currently not available on Windows based builds. This action is extremely useful when responding to both Brute Force and Denial of Service attacks in that, in both cases, you want to minimize both the network bandwidth and the data returned to the client. This action causes error message to appear in the log "(9)Bad file descriptor: core_output_filter: writing data to the network"
Description: Executes an external
script/binary supplied as parameter. As of v2.5.0, if the parameter
supplied to exec
is a Lua script (detected by the
.lua
extension) the script will be processed
internally. This means you will get direct access
to the internal request context from the script. Please read the
SecRuleScript
documentation for more details on how
to write Lua scripts.
Action Group: Non-disruptive
Example:
# The following is going to execute /usr/local/apache/bin/test.sh # as a shell script on rule match. SecRule REQUEST_URI "^/cgi-bin/script\.pl" \ "phase:2,t:none,t:lowercase,t:normalisePath,log,exec:/usr/local/apache/bin/test.sh" # The following is going to process /usr/local/apache/conf/exec.lua # internally as a Lua script on rule match. SecRule ARGS:p attack log,exec:/usr/local/apache/conf/exec.lua
The exec action is executed independently from any disruptive actions. External scripts will always be called with no parameters. Some transaction information will be placed in environment variables. All the usual CGI environment variables will be there. You should be aware that forking a threaded process results in all threads being replicated in the new process. Forking can therefore incur larger overhead in multi-threaded operation. The script you execute must write something (anything) to stdout. If it doesn't ModSecurity will assume execution didn't work.
Description: Configures a collection variable to expire after the given time in seconds.
Action Group: Non-disruptive
Example:
SecRule REQUEST_COOKIES:JSESSIONID "!^$" nolog,phase:1,pass,chain
SecAction setsid:%{REQUEST_COOKIES:JSESSIONID}
SecRule REQUEST_URI "^/cgi-bin/script\.pl" \
"phase:2,t:none,t:lowercase,t:normalisePath,log,allow,\
setvar:session.suspicious=1,expirevar:session.suspicious=3600,phase:1"
Note
You should use expirevar actions at the same time that you use setvar actions in order to keep the indented expiration time. If they are used on their own (perhaps in a SecAction directive) the expire time could get re-set. When variables are removed from collections, and there are no other changes, collections are not written to disk at the end of request. This is because the variables can always be expired again when the collection is read again on a subsequent request.
Description: Assigns a unique ID to the rule or chain.
Action Group: Meta-data
Example:
SecRule &REQUEST_HEADERS:Host "@eq 0" \
"log,id:60008,severity:2,msg:'Request Missing a Host Header'"
Note
These are the reserved ranges:
1-99,999; reserved for local (internal) use. Use as you see fit but do not use this range for rules that are distributed to others.
100,000-199,999; reserved for internal use of the engine, to assign to rules that do not have explicit IDs.
200,000-299,999; reserved for rules published at modsecurity.org.
300,000-399,999; reserved for rules published at gotroot.com.
400,000-419,999; unused (available for reservation).
420,000-429,999; reserved for ScallyWhack.
430,000-899,999; unused (available for reservation).
900,000-999,999; reserved for the Core Rules project.
1,000,000 and above; unused (available for reservation).
Description: Initialises a named persistent collection, either by loading data from storage or by creating a new collection in memory.
Action Group: Non-disruptive
Example: The following example initiates IP address tracking.
SecAction phase:1,initcol:ip=%{REMOTE_ADDR},nolog
Note
Normally you will want to use phase:1 along with initcol so that the collection is available in all phases.
Collections are loaded into memory when the initcol action is encountered. The collection in storage will be persisted (and the appropriate counters increased) only if it was changed during transaction processing.
See the "Persistant Storage" section for further details.
Description: Indicates that a successful match of the rule needs to be logged.
Action Group: Non-disruptive
Example:
SecAction phase:1,initcol:ip=%{REMOTE_ADDR},log
Note
This action will log matches to the Apache error log file and the ModSecurity audit log.
Description: Allows a data fragment to be logged as part of the alert message.
Action Group: Non-disruptive
Example:
SecRule &ARGS:p "@eq 0" "log,logdata:'%{TX.0}'"
Note
The logdata information appears in the error and/or audit log files and is not sent back to the client in response headers. Macro expansion is preformed so you may use variable names such as %{TX.0}, etc. The information is properly escaped for use with logging binary data.
Description: Assigns a custom message to the rule or chain.
Action Group: Meta-data
Example:
SecRule &REQUEST_HEADERS:Host "@eq 0" \ "log,id:60008,severity:2,msg:'Request Missing a Host Header'"
Note
The msg information appears in the error and/or audit log files and is not sent back to the client in response headers.
Description: If enabled ModSecurity will perform multiple operator invocations for every target, before and after every anti-evasion transformation is performed.
Action Group: Non-disruptive
Example:
SecDefaultAction log,deny,phase:1,t:removeNulls,t:lowercase
SecRule ARGS "attack" multiMatch
Note
Normally, variables are evaluated once, only after all transformation functions have completed. With multiMatch, variables are checked against the operator before and after every transformation function that changes the input.
Description: Indicates that a successful match of the rule should not be used as criteria whether the transaction should be logged to the audit log.
Action Group: Non-disruptive
Example:
SecRule REQUEST_HEADERS:User-Agent "Test" allow,noauditlog
Note
If the SecAuditEngine is set to On, all of the transactions will
be logged. If it is set to RelevantOnly, then you can control it with
the noauditlog action. Even if the noauditlog action is applied to a
specific rule and a rule either before or after triggered an audit
event, then the transaction will be logged to the audit log. The correct
way to disable audit logging for the entire transaction is to use
"ctl:auditEngine=Off
"
Description: Prevents rule matches from appearing in both the error and audit logs.
Action Group: Non-disruptive
Example:
SecRule REQUEST_HEADERS:User-Agent "Test" allow,nolog
Note
The nolog action also implies noauditlog.
Description: Continues processing with the next rule in spite of a successful match.
Action Group: Disruptive
Example1:
SecRule REQUEST_HEADERS:User-Agent "Test" log,pass
When using pass with SecRule with multiple targets, all targets will be processed and all non-disruptive actions will trigger for every match found. In the second example the TX:test target would be incremented by 1 for each matching argument.
Example2:
SecRule ARGS "test" log,pass,setvar:TX.test=+1
Note
The transaction will not be interrupted but a log will be generated for each matching target (unless logging has been suppressed).
Description: Pauses transaction processing for the specified number of milliseconds.
Action Group: Non-disruptive
Example:
SecRule REQUEST_HEADERS:User-Agent "Test" log,deny,status:403,pause:5000
Note
This feature can be of limited benefit for slowing down Brute Force Scanners, however use with care. If you are under a Denial of Service type of attack, the pause feature may make matters worse as this feature will cause child processes to sit idle until the pause is completed.
Description: Places the rule (or the rule chain) into one of five available processing phases.
Action Group: Meta-data
Example:
SecDefaultAction log,deny,phase:1,t:removeNulls,t:lowercase
SecRule REQUEST_HEADERS:User-Agent "Test" log,deny,status:403
Note
Keep in mind that is you specify the incorrect phase, the target variable that you specify may be empty. This could lead to a false negative situation where your variable and operator (RegEx) may be correct, but it misses malicious data because you specified the wrong phase.
Description: Prepends text given as parameter
to the response body. For this action to work content injection must be
enabled by setting SecContentInjection
to
On
. Also make sure you check the content type of the
response before you make changes to it (e.g. you don't want to inject
stuff into images).
Action Group: Non-disruptive
Processing Phases: 3 and 4.
Example:
SecRule RESPONSE_CONTENT_TYPE ^text/html "phase:3,nolog,pass,prepend:'Header<br>'"
While macro expansion is allowed in the additional content, you are strongly cautioned against inserting user defined data fields.
Description: Intercepts transaction by forwarding request to another web server using the proxy backend.
Action Group: Disruptive
Example:
SecRule REQUEST_HEADERS:User-Agent "Test" log,proxy:http://www.honeypothost.com/
Note
For this action to work, mod_proxy must also be installed. This action is useful if you would like to proxy matching requests onto a honeypot webserver.
Description: Intercepts transaction by issuing a redirect to the given location.
Action Group: Disruptive
Example:
SecRule REQUEST_HEADERS:User-Agent "Test" \
log,redirect:http://www.hostname.com/failed.html
Note
If the status
action is present
and its value is acceptable (301, 302, 303, or 307) it will be used for
the redirection. Otherwise status code 302 will be used.
Description: Specifies rule revision.
Action Group: Meta-data
Example:
SecRule REQUEST_METHOD "^PUT$" "id:340002,rev:1,severity:2,msg:'Restricted HTTP function'"
Note
This action is used in combination with the id
action to allow the same rule ID to be used
after changes take place but to still provide some indication the rule
changed.
Description: Sanitises (replaces each byte with an asterisk) a named request argument prior to audit logging.
Action Group: Non-disruptive
Example:
SecAction nolog,phase:2,sanitiseArg:password
Note
The sanitize actions do not sanitize any data within the actual raw requests but only on the copy of data within memory that is set to log to the audit log. It will not sanitize the data in the modsec_debug.log file (if the log level is set high enough to capture this data).
Description: Sanitises the variable (request argument, request header, or response header) that caused a rule match.
Action Group: Non-disruptive
Example: This action can be used to sanitise arbitrary transaction elements when they match a condition. For example, the example below will sanitise any argument that contains the word password in the name.
SecRule ARGS_NAMES password nolog,pass,sanitiseMatched
Note
Same note as sanitiseArg.
Description: Sanitises a named request header.
Action Group: Non-disruptive
Example: This will sanitise the data in the Authorization header.
SecAction log,phase:1,sanitiseRequestHeader:Authorization
Note
Same note as sanitiseArg.
Description: Sanitises a named response header.
Action Group: Non-disruptive
Example: This will sanitise the Set-Cookie data sent to the client.
SecAction log,phase:3,sanitiseResponseHeader:Set-Cookie
Note
Same note as sanitiseArg.
Description: Assigns severity to the rule it is placed with.
Action Group: Meta-data
Example:
SecRule REQUEST_METHOD "^PUT$" "id:340002,rev:1,severity:CRITICAL,msg:'Restricted HTTP function'"
Note
Severity values in ModSecurity follow those of syslog, as below:
0 - EMERGENCY
1 - ALERT
2 - CRITICAL
3 - ERROR
4 - WARNING
5 - NOTICE
6 - INFO
7 - DEBUG
It is possible to specify severity levels using either the numerical values or the text values. You should always specify severity levels using the text values. The use of the numerical values is deprecated (as of v2.5.0) and may be removed in one of the susequent major updates.
Description: Special-purpose action that
initialises the USER
collection.
Action Group: Non-disruptive
Example:
SecAction setuid:%{REMOTE_USER},nolog
Note
After initialisation takes place the variable USERID
will be available for use in the
subsequent rules.
Description: Special-purpose action that
initialises the SESSION
collection.
Action Group: Non-disruptive
Example:
# Initialise session variables using the session cookie value
SecRule REQUEST_COOKIES:PHPSESSID !^$ chain,nolog,pass
SecAction setsid:%{REQUEST_COOKIES.PHPSESSID}
Note
On first invocation of this action the collection will be empty
(not taking the predefined variables into account - see initcol
for more information). On subsequent
invocations the contents of the collection (session, in this case) will
be retrieved from storage. After initialisation takes place the
variable SESSIONID
will be available
for use in the subsequent rules.This action understands each application
maintains its own set of sessions. It will utilise the current web
application ID to create a session namespace.
Description: Creates, removes, or updates an environment variable.
Action Group: Non-disruptive
Examples:
To create a new variable (if you omit the value 1
will be used):
setenv:name=value
To remove a variable:
setenv:!name
Note
This action can be used to establish communication with other Apache modules.
Description: Creates, removes, or updates a variable in the specified collection.
Action Group: Non-disruptive
Examples:
To create a new variable:
setvar:tx.score=10
To remove a variable prefix the name with exclamation mark:
setvar:!tx.score
To increase or decrease variable value use +
and -
characters in front of a numerical value:
setvar:tx.score=+5
Description: Skips one or more rules (or chains) on successful match.
Action Group: Flow
Example:
SecRule REQUEST_URI "^/$" \
"phase:2,chain,t:none,skip:2"
SecRule REMOTE_ADDR "^127\.0\.0\.1$" "chain"
SecRule REQUEST_HEADERS:User-Agent "^Apache \(internal dummy connection\)$" "t:none"
SecRule &REQUEST_HEADERS:Host "@eq 0" \
"deny,log,status:400,id:960008,severity:4,msg:'Request Missing a Host Header'"
SecRule &REQUEST_HEADERS:Accept "@eq 0" \
"log,deny,log,status:400,id:960015,msg:'Request Missing an Accept Header'"
Note
Skip only applies to the current processing phase and not necessarily the order in which the rules appear in the configuration file. If you group rules by processing phases, then skip should work as expected. This action can not be used to skip rules within one chain. Accepts a single parameter denoting the number of rules (or chains) to skip.
Description: Skips rules (or chains) on
successful match resuming rule execution after the specified rule ID or
marker (see SecMarker
) is found.
Action Group: Flow
Example:
SecRule REQUEST_URI "^/$" "chain,t:none,skipAfter:960015"
SecRule REMOTE_ADDR "^127\.0\.0\.1$" "chain"
SecRule REQUEST_HEADERS:User-Agent "^Apache \(internal dummy connection\)$" "t:none"
SecRule &REQUEST_HEADERS:Host "@eq 0" \
"deny,log,status:400,id:960008,severity:4,msg:'Request Missing a Host Header'"
SecRule &REQUEST_HEADERS:Accept "@eq 0" \
"log,deny,log,status:400,id:960015,msg:'Request Missing an Accept Header'"
Note
SkipAfter
only applies to the current
processing phase and not necessarily the order in which the rules appear
in the configuration file. If you group rules by processing phases, then
skip should work as expected. This action can not be used to skip rules
within one chain. Accepts a single parameter denoting the last rule ID
to skip.
Description: Specifies the response status
code to use with actions deny
and redirect
.
Action Group: Data
Example:
SecDefaultAction log,deny,status:403,phase:1
Note
Status actions defined in Apache scope locations (such as Directory, Location, etc...) may be superseded by phase:1 action settings. The Apache ErrorDocument directive will be triggered if present in the configuration. Therefore if you have previously defined a custom error page for a given status then it will be executed and its output presented to the user.
Description: This action can be used which transformation function should be used against the specified variables before they (or the results, rather) are run against the operator specified in the rule.
Action Group: Non-disruptive
Example:
SecDefaultAction log,deny,phase:1,t:removeNulls,t:lowercase
SecRule REQUEST_COOKIES:SESSIONID "47414e81cbbef3cf8366e84eeacba091" \
log,deny,status:403,t:md5,t:hexEncode
Note
Any transformation functions that you specify in a SecRule will be in addition to previous ones specified in SecDefaultAction. Use of "t:none" will remove all transformation functions for the specified rule.
Description: Assigns custom text to a rule or chain.
Action Group: Meta-data
Example:
SecRule REQUEST_FILENAME "\b(?:n(?:map|et|c)|w(?:guest|sh)|cmd(?:32)?|telnet|rcmd|ftp)\.exe\b" \
"t:none,t:lowercase,deny,msg:'System Command Access',id:'950002',\
tag:'WEB_ATTACK/FILE_INJECTION',tag:'OWASP/A2',severity:'2'"
Note
The tag information appears in the error and/or audit log files. Its intent is to be used to automate classification of rules and the alerts generated by rules. Multiple tags can be used per rule/chain.
Description: This action should be used together with an XPath expression to register a namespace.
Action Group: Data
Example:
SecRule REQUEST_HEADERS:Content-Type "text/xml" \
"phase:1,pass,ctl:requestBodyProcessor=XML,ctl:requestBodyAccess=On, \
xmlns:xsd="http://www.w3.org/2001/XMLSchema"
SecRule XML:/soap:Envelope/soap:Body/q1:getInput/id() "123" phase:2,deny
A number of operators can be used in rules, as documented below. The
operator syntax uses the @
symbol followed by the
specific operator name.
Description: This operator is a string
comparison and returns true if the parameter value is found at the
beginning of the input. Macro expansion is performed so you may use
variable names such as %{TX.1}
, etc.
Example:
SecRule REQUEST_LINE "!@beginsWith GET" t:none,deny,status:403 SecRule REQUEST_ADDR "^(.*)\.\d+$" deny,status:403,capture,chain SecRule ARGS:gw "!@beginsWith %{TX.1}"
Description: This operator is a string comparison and returns true if the parameter value is found anywhere in the input. Macro expansion is performed so you may use variable names such as %{TX.1}, etc.
Example:
SecRule REQUEST_LINE "!@contains .php" t:none,deny,status:403 SecRule REQUEST_ADDR "^(.*)$" deny,status:403,capture,chain SecRule ARGS:ip "!@contains %{TX.1}"
Description: This operator is a string comparison and returns true if the parameter value is found at the end of the input. Macro expansion is performed so you may use variable names such as %{TX.1}, etc.
Example:
SecRule REQUEST_LINE "!@endsWith HTTP/1.1" t:none,deny,status:403 SecRule ARGS:route "!@endsWith %{REQUEST_ADDR}" t:none,deny,status:403
Description: This operator is a numerical comparison and stands for "equal to."
Example:
SecRule &REQUEST_HEADERS_NAMES "@eq 15"
Description: This operator is a numerical comparison and stands for "greater than or equal to."
Example:
SecRule &REQUEST_HEADERS_NAMES "@ge 15"
Description: This operator looks up various
data fields from an IP address or hostname in the target data. The
results will be captured in the GEO
collection.
You must provide a database via SecGeoLookupDb
before this operator can be
used.
This operator matches and the action is executed on a
successful lookup. For this reason, you probably want to
use the pass,nolog actions. This allows for
setvar
and other non-disruptive
actions to be executed on a match. If you wish to block on a failed
lookup, then do something like this (look for an empty GEO
collection):
SecGeoLookupDb /usr/local/geo/data/GeoLiteCity.dat ... SecRule REMOTE_ADDR "@geoLookup" "pass,nolog" SecRule &GEO "@eq 0" "deny,status:403,msg:'Failed to lookup IP'"
See the GEO
variable for an
example and more information on various fields available.
Description: This operator is a numerical comparison and stands for "greater than."
Example:
SecRule &REQUEST_HEADERS_NAMES "@gt 15"
Description: Executes the external script/binary given as parameter to the operator against every file extracted from the request. As of v2.5.0, if the supplied filename is not absolute it is treated as relative to the directory in which the configuration file resides. Also as of v2.5.0, if the filename is determined to be a Lua script (based on its extension) the script will be processed by the internal engine. As such it will have full access to the ModSecurity context.
Example of using an external binary/script:
# Execute external script to validate uploaded files.
SecRule FILES_TMPNAMES "@inspectFile /opt/apache/bin/inspect_script.pl"
Example of using Lua script:
SecRule FILES_TMPNANMES "@inspectFile inspect.lua"
Script inspect.lua
:
function main(filename) -- Do something to the file to verify it. In this example, we -- read up to 10 characters from the beginning of the file. local f = io.open(filename, "rb"); local d = f:read(10); f:close(); -- Return null if there is no reason to believe there is ansything -- wrong with the file (no match). Returning any text will be taken -- to mean a match should be trigerred. return null; end
Description: This operator is a numerical comparison and stands for "less than or equal to."
Example:
SecRule &REQUEST_HEADERS_NAMES "@le 15"
Description: This operator is a numerical comparison and stands for "less than."
Example:
SecRule &REQUEST_HEADERS_NAMES "@lt 15"
Description: Phrase Match operator. This operator uses a set based matching engine (Aho-Corasick) for faster matches of keyword lists. It will match any one of its arguments anywhere in the target value. The match is case insensitive.
Example:
SecRule REQUEST_HEADERS:User-Agent "@pm WebZIP WebCopier Webster WebStripper SiteSnagger ProWebWalker CheeseBot" "deny,status:403
The above would deny access with 403 if any of the words matched within the User-Agent HTTP header value.
Description: Phrase Match operator. This
operator uses a set based matching engine (Aho-Corasick) for faster
matches of keyword lists. This operator is the same as
@pm
except that it takes a list of files as
arguments. It will match any one of the phrases listed in the file(s)
anywhere in the target value.
Notes:
The contents of the files should be one phrase per line. End of line markers will be stripped from the phrases, however, whitespace will not be trimmed from phrases in the file. Empty lines and comment lines (beginning with a '#') are ignored.
To allow easier inclusion of phrase files with rulesets, relative paths may be used to the phrase files. In this case, the path of the file containing the rule is prepended to the phrase file path.
Example:
SecRule REQUEST_HEADERS:User-Agent "@pm /path/to/blacklist1 blacklist2" "deny,status:403
The above would deny access with 403 if any of the patterns in the
two files matched within the User-Agent HTTP header value. The
blacklist2
file would need to be placed in the same
path as the file containing the rule.
Description: Look up the parameter in the RBL given as parameter. Parameter can be an IPv4 address, or a hostname.
Example:
SecRule REMOTE_ADDR "@rbl sc.surbl.org"
Description: Regular expression operator. This is the default operator, so if the "@" operator is not defined, it is assumed to be rx.
Example:
SecRule REQUEST_HEADERS:User-Agent "@rx nikto"
Note
Regular expressions are handled by the PCRE library (http://www.pcre.org). ModSecurity compiles its regular expressions with the following settings:
The entire input is treated as a single line, even when there are newline characters present.
All matches are case-sensitive. If you do not care about case
sensitivity you either need to implement the lowercase
transformation function, or use
the per-pattern(?i)
modifier, as
allowed by PCRE.
The PCRE_DOTALL
and
PCRE_DOLLAR_ENDONLY
flags are set
during compilation, meaning a single dot will match any character,
including the newlines and a $
end anchor will not match a trailing newline character.
Description: This operator is a string comparison and returns true if the parameter value matches the input exactly. Macro expansion is performed so you may use variable names such as %{TX.1}, etc.
Example:
SecRule ARGS:foo "!@streq bar" t:none,deny,status:403 SecRule REQUEST_ADDR "^(.*)$" deny,status:403,capture,chain SecRule REQUEST_HEADERS:Ip-Address "!@streq %{TX.1}"
Description: Validates the byte range used in the variable falls into the specified range.
Example:
SecRule ARGS:text "@validateByteRange 10, 13, 32-126"
Note
You can force requests to consist only of bytes from a certain
byte range. This can be useful to avoid stack overflow attacks (since
they usually contain "random" binary content). Default range values are
0 and 255, i.e. all byte values are allowed. This directive does not
check byte range in a POST payload when
multipart/form-data
encoding (file upload) is used.
Doing so would prevent binary files from being uploaded. However, after
the parameters are extracted from such request they are checked for a
valid range.
validateByteRange is similar to the ModSecurity 1.X SecFilterForceByteRange Directive however since it works in a rule context, it has the following differences:
You can specify a different range for different variables.
It has an "event" context (id, msg....)
It is executed in the flow of rules rather than being a built in pre-check.
Description: Validates the DOM tree generated by the XML request body processor against the supplied DTD.
Example:
SecDefaultAction log,deny,status:403,phase:2
SecRule REQUEST_HEADERS:Content-Type ^text/xml$ \
phase:1,t:lowercase,nolog,pass,ctl:requestBodyProcessor=XML
SecRule REQBODY_PROCESSOR "!^XML$" nolog,pass,skipAfter:12345
SecRule XML "@validateDTD /path/to/apache2/conf/xml.dtd" "deny,id:12345"
This operator requires request body to be processed as XML.
Description: Validates the DOM tree generated by the XML request body processor against the supplied XML Schema.
Example:
SecDefaultAction log,deny,status:403,phase:2
SecRule REQUEST_HEADERS:Content-Type ^text/xml$ \
phase:1,t:lowercase,nolog,pass,ctl:requestBodyProcessor=XML
SecRule REQBODY_PROCESSOR "!^XML$" nolog,pass,skipAfter:12345
SecRule XML "@validateSchema /path/to/apache2/conf/xml.xsd" "deny,id:12345"
This operator requires request body to be processed as XML.
Description: Verifies the encodings used in the variable (if any) are valid.
Example:
SecRule ARGS "@validateUrlEncoding"
Note
URL encoding is an HTTP standard for encoding byte values within a
URL. The byte is escaped with a % followed by two hexadecimal values
(0-F). This directive does not check encoding in a POST payload when the
multipart/form-data
encoding (file upload) is used.
It is not necessary to do so because URL encoding is not used for this
encoding.
Description: Verifies the variable is a valid UTF-8 encoded string.
Example:
SecRule ARGS "@validateUtf8Encoding"
Note
UTF-8 encoding is valid on most web servers. Integer values between 0-65535 are encoded in a UTF-8 byte sequence that is escaped by percents. The short form is two bytes in length.
check for three types of errors:
Not enough bytes. UTF-8 supports two, three, four, five, and six byte encodings. ModSecurity will locate cases when a byte or more is missing.
Invalid encoding. The two most significant bits in most characters are supposed to be fixed to 0x80. Attackers can use this to subvert Unicode decoders.
Overlong characters. ASCII characters are mapped directly into the Unicode space and are thus represented with a single byte. However, most ASCII characters can also be encoded with two, three, four, five, and six characters thus tricking the decoder into thinking that the character is something else (and, presumably, avoiding the security check).
Description: This operator verifies a given regular expression as a potential credit card number. It first matches with a single generic regular expression then runs the resulting match through a Luhn checksum algorithm to further verify it as a potential credit card number.
Example:
SecRule ARGS "@verifyCC \d{13,16}" \
"phase:2,sanitiseMatched,log,auditlog,pass,msg:'Potential credit card number'"
Description: This operator is a string
comparison and returns true if the input value is found anywhere within
the parameter value. Note that this is similar to
@contains
, except that the target and match values
are reversed. Macro expansion is performed so you may use variable names
such as %{TX.1}, etc.
Example:
SecRule REQUEST_METHOD "!@within get,post,head" t:lowercase,deny,status:403 SecAction "pass,setvar:'tx.allowed_methods=get,post,head'" SecRule REQUEST_METHOD "!@within %{tx.allowed_methods}" t:lowercase,deny,status:403
Macros allow for using place holders in rules that will be expanded out to their values at runtime. Currently only variable expansion is supported, however more options may be added in future versions of ModSecurity.
Format:
%{VARIABLE} %{COLLECTION.VARIABLE}
Macro expansion can be used in actions such as initcol, setsid, setuid, setvar, setenv, logdata. Operators that are evaluated at runtime support expansion and are noted above. Such operators include @beginsWith, @endsWith, @contains, @within and @streq. You cannot use macro expansion for operators that are "compiled" such as @pm, @rx, etc. as these operators have their values fixed at configure time for efficiency.
Some values you may want to expand include: TX, REMOTE_ADDR, USERID, HIGHEST_SEVERITY, MATCHED_VAR, MATCHED_VAR_NAME, MULTIPART_STRICT_ERROR, RULE, SESSION, USERID, among others.
At this time it is only possible to have three collections in which
data is stored persistantly (i.e. data available to multiple requests).
These are: IP
, SESSION
and USER
.
Every collection contains several built-in variables that are available and are read-only unless otherwise specified:
CREATE_TIME
- date/time of
the creation of the collection.
IS_NEW
- set to 1 if the
collection is new (not yet persisted) otherwise set to 0.
KEY
- the value of the
initcol variable (the client's IP address in the example).
LAST_UPDATE_TIME
- date/time
of the last update to the collection.
TIMEOUT
- date/time in
seconds when the collection will be updated on disk from memory (if no
other updates occur). This variable may be set if you wish to specifiy
an explicit expiration time (default is 3600 seconds).
UPDATE_COUNTER
- how many
times the collection has been updated since creation.
UPDATE_RATE
- is the average
rate updates per minute since creation.
To create a collection to hold session variables (SESSION
) use action setsid
. To create a collection to hold user
variables (USER
) use action setuid
. To create a collection to hold client
address variables (IP
) use action
initcol
.
ModSecurity implements atomic updates of persistent variables only
for integer variables (counters) at this time. Variables are read from
storage whenever initcol
is encountered in the rules
and persisted at the end of request processing. Counters are adjusted by
applying a delta generated by re-reading the persisted data just before
being persisted. This keeps counter data consistent even if the counter
was modified and persisted by another thread/process during the
transaction.
ModSecurity uses a Berkley Database (SDBM) for persistant storage. This type of database is generally limited to storing a maximum of 1008 bytes per key. This may be a limitation if you are attempting to store a considerable amount of data in variables for a single key. Some of this limitation is planned to be reduced in a future version of ModSecurity.
Web application firewalls have a difficult job trying to make sense of data that passes by, without any knowledge of the application and its business logic. The protection they provide comes from having an independent layer of security on the outside. Because data validation is done twice, security can be increased without having to touch the application. In some cases, however, the fact that everything is done twice brings problems. Problems can arise in the areas where the communication protocols are not well specified, or where either the device or the application do things that are not in the specification. In such cases it may be possible to design payload that will be interpreted in one way by one device and in another by the other device. This problem is better known as Impedance Mismatch. It can be exploited to evade the security devices.
While we will continue to enhance ModSecurity to deal with various evasion techniques the problem can only be minimized, but never solved. With so many different application backend chances are some will always do something completely unexpected. The only solution is to be aware of the technologies in the backend when writing rules, adapting the rules to remove the mismatch. See the next section for some examples.
When writing rules to protect PHP applications you need to pay attention to the following facts:
When "register_globals" is set to "On" request parameters are automatically converted to script variables. In some PHP versions it is even possible to override the $GLOBALS array.
Whitespace at the beginning of parameter names is ignored. (This is very dangerous if you are writing rules to target specific named variables.)
The remaining whitespace (in parameter names) is converted to underscores. The same applies to dots and to a "[" if the variable name does not contain a matching closing bracket. (Meaning that if you want to exploit a script through a variable that contains an underscore in the name you can send a parameter with a whitespace or a dot instead.)
Cookies can be treated as request parameters.
The discussion about variable names applies equally to the cookie names.
The order in which parameters are taken from the request and the environment is EGPCS (environment, GET, POST, Cookies, built-in variables). This means that a POST parameter will overwrite the parameters transported on the request line (in QUERY_STRING).
When "magic_quotes_gpc" is set to "On" PHP will use backslash to escape the following characters: single quote, double quote, backslash, and the nul byte.
If "magic_quotes_sybase" is set to "On" only the single quote will be escaped using another single quote. In this case the "magic_quotes_gpc" setting becomes irrelevant. The "magic_quotes_sybase" setting completely overrides the "magic_quotes_gpc" behaviour but "magic_quotes_gpc" still must be set to "On" for the Sybase-specific quoting to be work.
PHP will also automatically create nested arrays for you. For example "p[x][y]=1" results in a total of three variables.