WordPress.org

Data Validation

Languages: English • 日本語 • (Add your language)

Untrusted data comes from many sources (users, third party sites, your own database!, ...) and all of it needs to be validated both on input and output.

Output Sanitation

The method of data sanitation depends on the type of data and the context in which it is used. Below are some common tasks in WordPress and how they should be sanitized.

Integers

intval( $int ) or (int) $int: If it's supposed to be an integer, cast it as one.
absint( $int ): Ensures that the result is nonnegative.

HTML/XML

Note that many types of XML documents (as opposed to HTML documents) understand only a few named character references: apos, amp, gt, lt, quot. When outputting text to such an XML document, be sure to filter any text containing illegal named entities through WordPress's ent2ncr( $text ) function.

HTML/XML Fragments

wp_kses( (string) $fragment, (array) $allowed_html, (array) $protocols = null ): KSES Strips Evil Scripts. All untrusted HTML (post text, comment text, etc.) should be run through wp_kses(). See wp-includes/kses.php for documentation, defaults, etc.
wp_rel_nofollow( (string) $html ): Adds a "rel='nofollow'" attribute to any <a> link.

Text Nodes

esc_html( $text ) (since 2.8): Encodes < > & " ' (less than, greater than, ampersand, double quote, single quote). Very similar to esc_attr.
esc_html__ (since 2.8): Translates and encodes
esc_html_e (since 2.8): Translates, encodes, and echos
wp_specialchars( $string, $quote_style = ENT_NOQUOTES, $charset = false, $double_encode = false ) (deprecated since 2.8): Encodes < > & (less than, greater than, ampersand). Will never double encode entities. Since 2.8, if called with exactly 1 argument, it will encode quote characters as well (via esc_html), as extra protection for older plugins.
htmlspecialchars( $text, ENT_NOQUOTES ): Encodes < > &. Will double encode html entities if run twice.

Attribute Nodes

esc_attr (since 2.8)
attribute_escape( $text ) (deprecated since 2.8): Encodes < > & " ' (less than, greater than, ampersand, double quote, single quote). Will never double encode entities. See clean_url() in #URLs
esc_attr__: Translates and encodes
esc_attr_e: Translates, encodes, and echos
htmlspecialchars( $text, ENT_QUOTES ): Encodes < > & " '. Will double encode html entities if run twice. See clean_url() in #URLs

JavaScript

esc_js (since 2.8)
js_escape( $text ) (deprecated since 2.8): Escapes ', encodes ", and fixes line endings.

URLs

esc_url (since 2.8)

clean_url( $url, (array) $protocols = null, $context = 'display' )

Always use clean_url when sanitizing URLs (in text nodes, attribute nodes or anywhere else). Rejects URLs that do not have one of the provided whitelisted protocols (defaulting to http, https, ftp, ftps, mailto, news, irc, gopher, nntp, feed, and telnet), eliminates invalid characters, and removes dangerous characters. $context should be one of the following.

display: For output in an (X)HTML or XML document. Encodes ampersands (&) and single quotes (') as numeric entity references (&#038, &#039).
url: strips invalid URL characters only.
db: use if you're about to insert into the database.

esc_url_raw (since 2.8): For inserting an URL in the database (just like $context = "db", above)
urlencode( $scalar ): Encodes for use in URL (as a query parameter, for example)
urlencode_deep( $array ): urlencodes all array elements.

Database

$wpdb->insert( $table, (array) $data ): $data should be unescaped (the function will escape them for you). Keys are columns, Values are values.
$wpdb->update( $table, (array) $data, (array) $where ): $data should be unescaped. Keys are columns, Values are values. $where should be unescaped. Multiple WHERE conditions are ANDed together.

$wpdb->update(
  'my_table',
  array( 'status' => $untrusted_status, 'title' => $untrusted_title ),
  array( 'id' => 123 )
);

$wpdb->prepare( $format, (scalar) $value1, (scalar) $value2, ... ): $format is a sprintf() like format string. It only understands %s and %d, neither of which needs to be enclosed in quotation marks.

$wpdb->get_var( $wpdb->prepare(
  "SELECT something FROM table WHERE foo = %s and status = %d",
  $name, // an unescaped string (function will do the sanitation for you)
  $status // an untrusted integer (function will do the sanitation for you)
) );

esc_sql( $text ) (since 2.8)
$wpdb->escape( $text ): Escapes a single string for use in a SQL query. Glorified addslashes().
$wpdb->escape_by_ref( &$text ): No return value.
like_escape( $string ): Sanitizes $string for use in a LIKE expression of a SQL query. Will still need to be SQL escaped (with one of the above functions).

Filesystem

validate_file( (string) $filename, (array) $allowed_files = "" ): Used to prevent directory traversal attacks, or to test a filename against a whitelist. Returns 0 if $filename represents a valid relative path. After validating, you must treat $filename as a relative path (i.e. you must prepend it with an absolute path), since something like /etc/hosts will validate with this function. Returns an integer greater than zero if the given path contains .., ./, or :, or is not in the $allowed_files whitelist. Be careful making boolean interpretations of the result, since false (0) indicates the filename has passed validation, whereas true (> 0) indicates failure.

HTTP Headers

Header splitting attacks are annoying since they are dependent on the HTTP client. WordPress has little need to include user generated content in HTTP headers, but when it does, WordPress typically uses whitelisting for most of its HTTP headers.

WordPress does use user generated content in HTTP Location headers, and provides sanitation for those.

wp_redirect($location, $status = 302): A safe way to redirect to any URL. Ensures the resulting HTTP Location header is legitimate.
wp_safe_redirect($location, $status = 302): Even safer. Only allows redirects to whitelisted domains.

Input Validation

Many of the functions above in #Output_Sanitation are useful for input validation. In addition, WordPress uses the following functions.

Slugs

sanitize_title( $title ): Used in post slugs, for examlpe
sanitize_user( $username, $strict = false ): Use $strict when creating a new user (though you should use the API for that).

HTML

balanceTags( $html ) or force_balance_tags( $html ): Tries to make sure HTML tags are balanced so that valid XML is output.
tag_escape( $html_tag_name ): Sanitizes an HTML tag name (does not escape anything, despite the name of the function).

Email

is_email( $email_address ): returns boolean

Arrays

array_map( 'absint', $array ): Ensures all elements are nonnegative integers. Replace callback with whatever is appropriate for your data.

Validation Philosophies

There are several different philosophies about how validation should be done. Each is appropriate for different scenarios.

Whitelist

Accept data only from a finite list of known and trusted values.

$possible_values = array( 'a', 1, 'good' );
if ( !in_array( $untrusted, $possible_values ) )
  die( "Don't do that!" );

// Be careful here with fancy breaks and default actions.
switch ( $untrusted ) {
case 'a' :
  ...
  break;
...
default :
  die( "You hoser!" );
}

Blacklist

Reject data from finite list of known untrusted values. This is very rarely a good idea.

Format Detection

Test to see if the data is of the correct format. Only accept it if it is.

if ( !ctype_alnum( $data ) )
  die( "Your data is teh suX0R" );
if ( preg_match( "/[^0-9.-]/", $data ) )
  die( "Float on somewhere else, jerky" );

Format Correction

Accept most any data, but remove or alter the dangerous pieces.

$trusted_integer = (int) $untrusted_integer;
$trusted_alpha = preg_replace( '/[^a-z]/i', "", $untrusted_alpha );
$trusted_slug = sanitize_title( $untrusted_slug );

Changelog

2.8: Deprecated the following functions. (via WordPress Development Updates)
- clean_url() -> esc_url()
- sanitize_url() -> esc_url_raw()
- wp_specialchars() -> esc_html() (also: esc_html__() and esc_html_e())
- attribute_escape() -> esc_attr() (also: esc_attr__() and esc_attr_e())