ExpressionEngine Docs

Security Guidelines

Cross Site Scripting (XSS)

Cross Site Scripting is a type of security vulnerability that allows code injection by malicious users onto a page. You can find some educational reading and examples on the following site:

Cross Site Scripting should be taken very seriously as you would never want your add-on to be the source of an attack vector.


ee()->security->xss_clean() is the built in ExpressionEngine XSS sanitization method, which is constantly tweaked for improved security and performance. It accepts both a string and an array and will return sanitized text:

$str = ee()->security->xss_clean($str);

Sanitized Variables on Input

Keys are converted to UTF-8 and new lines in data are normalized automatically by the Input Class for the following variables:

However, for performance reasons, data are not automatically run through the xss filter. If you are storing or displaying data from these variables, you should use the Input class’s get(), post(), or get_post() methods and pass TRUE as the second parameter so the value will be XSS cleaned.

If the user input is in the Control Panel, such as a module’s back end, it is at your discretion based on the needs of the module for whether or not administrative input is sanitized. Always err on the side of caution, and never assume that your end user will only allow access to the back end of your module to trusted members.


As ExpressionEngine assumes that all information in the database is sanitized against XSS, the responsibility for sanitization falls on input to the database. Active Record methods will escape your data, but will not XSS clean it.Therefore, all data should be run through either ee()->security->xss_clean() or ee()->input->get_post('key', TRUE) before being stored in the database.

Outputting Data to the Page

Use the Typography class whenever outputting blocks of content from user submitted data. It is regularly updated to improve security and performance, saving you both time and energy.

When In Doubt…

If there is any doubt on the safety of a variable that you are outputting or inserting into the database, use XSS Clean: ee()->security->xss_clean($value).

SQL Injection Prevention

SQL Injection is a special type of attack in which data is used in a query without being properly filtered, allowing a user to execute their own queries on the database. Example:

$evil = "brett'; DELETE FROM exp_members;";
$query = ee()->db->query("SELECT * FROM exp_members WHERE username='{$evil}'");

For more information, you can read MySQL’s guide to SQL Injection security:

Escaping PHP Variables

PHP variables should be escaped in queries anytime the variable is not explicitly set to a hard-coded value within the method using the query. This means that even variables passed as arguments to a method must be escaped before being used in a query.

Manually written queries should use both XSS cleaned data and ee()->db->escape_str() on variables, even if you think the value is trusted:

$data = ee()->security->xss_clean($foo);


$data = ee()->input->get_post('foo', TRUE);


$query = ee()->db->query("SELECT field FROM table WHERE column = '".ee()->db->escape_str($data)."'");

ee()->db->insert() is the preferred method for INSERT queries, as values are escaped automatically in the supplied data array:

        'name'          => 'Brett Bretterson',
        'email_address' => ''

ee()->db->update() is the preferred method for UPDATE queries, as values are escaped automatically in the supplied data and where arrays:

    array('email_address' => ''),
    array('name' => 'Brett Bretterson')

Note: If you send the third argument (the WHERE clause) as an array as shown above, it will automatically be escaped. If you send a string, you must escape it yourself:

    array('email_address' => ''),
    "name = '".ee()->db->escape_str($foo)."'"

Tag Parameters

Never Assume Tag Parameters are “Good” Input

Do not make security exceptions for tag parameters. With PHP on Input, nested tags, other plugins, or variables being possible sources for parameter values, you cannot be sure that the data is safe.

Validate Values Before Using

Always validate the values being supplied to a tag parameter before using them in your code. switch() statements are good for numerous possible values, as are arrays of possible values:

switch ($foo = ee()->TMPL->fetch_param('foo'))
    case 'bar':
    case 'baz':
    case 'bag':
        // value is already set, and okay, so simply break
        $foo = '';

$valid_foo = array('bar', 'baz', 'bag');
$foo = (in_array($foo = ee()->TMPL->fetch_param('foo'), $valid_foo)) ? $foo : '';

If you cannot validate against specific values, at least validate the type of data:

if (! ctype_digit($foo = ee()->TMPL->fetch_param('foo')))
    ee()->TMPL->log_item('Super Class Module error: Provided parameter "foo" contains non-digit characters');
    return FALSE;

Or even:

$foo = (ctype_digit($foo = ee()->TMPL->fetch_param('foo'))) ? FALSE : $foo;

Note: You no doubt notice that ctype_digit is being used here to validate the parameter as a numeric value. Why? is_numeric() returns TRUE for some non-integer numbers, including notation, e.g. “-0123.45e6”. is_int() only returns TRUE on actual integer variable types, and tag parameters are always strings. Note that ctype_digit(), will return TRUE on an empty string in pre-5.1.0 versions of PHP.

Default Values

Always have default values if you plan to allow the code to execute without parameters being supplied, or in the case of invalid parameter values being provided. An empty string, NULL, or boolean FALSE simply needs to be tested later to accommodate defaults in your code. This also allows you to change the defaults all in one place in the script. Here is one method, that takes advantage of PHP’s variable variables.

$defaults = array(
    'type'    => '',
    'show_foo'  => FALSE,
    'limit'   => 5

foreach ($defaults as $key => $val)
    $key = ($key = ee()->TMPL->fetch_param($key)) ? $key : $val;

// Results in three variables being set:
// $type, $show_foo, and $limit, to their corresponding tag parameter value
// or the default value if the parameter was not present
// Each variable would still need to be validated as instructed above
// before using them in the code.

Cross Site Request Forgery

To help prevent spam and protect against Cross-site Request Forgery (CSRF), ExpressionEngine adds a random string to a hidden field on all forms. A copy of this string - also know as a CSRF token - is stored in the database along with the session id for which that form was generated. When the form is submitted this field is checked before any processing is done. If no CSRF token is present or no match is found, then the submission is rejected.

CSRF Tokens in Templates

If you are manually creating templates that send POST requests you must include the CSRF token as part of the form. This is easily done using the csrf_token variable as a value for a hidden field called csrf_token:

<input type="hidden" name="csrf_token" value="{csrf_token}">

Creating Template Forms from Add-ons

If your add-on is creating a form for the template, you should use ee()->functions->form_declaration() <form_declaration>. This automatically adds the CSRF token as a hidden input field. It also allows any extensions the site may have installed to modify the form before it is served, thus creating a more uniform experience for the end user. :

  'action'  => ''

If your form submits to a different site you should ensure that you are not leaking the user’s CSRF token. You can either do this by manually creating the form open tag or setting the ‘secure’ option for the form_declaration() method to FALSE. :

  'secure'  => FALSE

Validating Form Hashes in Your Add-on

ExpressionEngine will automatically check the CSRF token of all requests before handing the request off to your add-on. This means that all forms and requests must include the csrf_token field. Asynchronous requests that include an HTTP_REQUESTED_BY header (this is set by most popular libraries, such as jQuery) default to being exempt from these checks as they provide a good layer of intrinsic security.

There are several ways in which you can control this validation behavior of the CSRF tokens.

Disabling the check

For action requests you can disable all CSRF token checks. This is done by setting the csrf_exempt column in the actions table to 1 for that action.

You should only do this for actions that do not add, delete, or otherwise modify data (e.g. search) or requests that are expect to be initiated by another site (e.g. webhooks, payment gateways, etc).

Forcing AJAX Validation

While the same origin restriction for AJAX requests provides a good level of security from cross-site request forgery, compromised browser add-ons can send these requests.

If you have AJAX action requests that are performing sensitive operations, then you should consider forcing AJAX CSRF validation for your add-on. This happens on per-class basis using a marker interface. You simply implement the Strict_XID interface on your action receiving class:

class My_module implements Strict_XID { ... }

You can still disable the check on a per-action basis.

Forms in the Control Panel

The Control Panel’s Display class automatically adds hashes to any form using the form_open() helper. CSRF tokens are a requirement in the Control Panel and as such the check cannot be disabled. The Control Panel includes a jQuery ajax prefilter that takes care of CSRF tokens on all AJAX requests and also handles periodic token refreshing for additional security.

You should use EE.CSRF_TOKEN if you require the token in your JavaScript. Due to the ephemeral nature of CSRF tokens you should access this property when you need it. It should not be copied or cached.

Handling Form Submissions

Form submissions are the most common form of user input you will handle in your add-ons, so it is important to understand how to deal with them securely.

Outputting Form Data to the Screen

Never output unfiltered incoming data directly to the screen.

Trust No One

Treat all input as potentially dangerous, even from within the control panel.

Use a Logic Map for Processing

In your methods that will be handling form data, create a logic map that you can use to ensure that you are handling all validation and security checks prior to performing any actions. The following list contains common things to use; your add-on may have fewer or additional requirements.

After processing, make sure submitted data that might be sent to the screen for a success or error message is the filtered and validated version

Filename Security

include() and require()

Many servers have the ability to include files from offsite or anywhere in the local server, so when using include() or require() with user submitted data you need to be extremely careful. The best practice is to not design your add-on in such a way that would make this necessary in the first place, but if you do, either:

Saving Images or Files to the Server

When saving images or files to the server, make sure and validate the file type (MIME) and also clean the file name to remove possible naughty characters.

Move most of the preferences and settings to an add-on guidelines page

Preferences and Settings

Storage of Settings

Security and required preference settings should be stored in the database or config.php file.

Use of Settings in Forms

Never send values for preferences or settings in hidden form fields. HTML source is open and readable, so a malicious user could simply copy the HTML or use a browser plugin to alter the form data to something you do not expect or desire. If absolutely required, encode them:

Yes / No Preferences

If your preference setting is a simple Yes / No, use 'y' for Yes and 'n' for No in both the code and the database, to keep things simple and consistent.

Follow the Art of KISS

“Keep It Simple, Stupid”. Before adding a preference, ask yourself: is a preference for ‘foo’ really needed? Eventually with too many preferences, there will be interference and priority issues, and over-complication.

General Security Practice