XSS filtering in PHP (cleans various UTF encodings & nested exploits)

xss_clean.php

<?php
/*
 * XSS filter, recursively handles HTML tags & UTF encoding
 * Optionally handles base64 encoding
 *
 * ***DEPRECATION RECOMMENDED*** Not updated or maintained since 2011
 * A MAINTAINED & BETTER ALTERNATIVE => kses
 * https://github.com/RichardVasquez/kses/
 *
 * This was built from numerous sources
 * (thanks all, sorry I didn't track to credit you)
 *
 * It was tested against *most* exploits here: http://ha.ckers.org/xss.html
 * WARNING: Some weren't tested!!!
 * Those include the Actionscript and SSI samples, or any newer than Jan 2011
 *
 */

class xssClean {

    /*
     * Recursive worker to strip risky elements
     *
     * @param   string  $input      Content to be cleaned. It MAY be modified in output
     * @return  string  $output     Modified $input string
     */
    public function clean_input( $input, $safe_level = 0 ) {

        $output = $input;
        do {
            // Treat $input as buffer on each loop, faster than new var
            $input = $output;
            
            // Remove unwanted tags
            $output = $this->strip_tags( $input );
            $output = $this->strip_encoded_entities( $output );

            // Use 2nd input param if not empty or '0'
            if ( $safe_level !== 0 ) {
                $output = $this->strip_base64( $output );
            }

        } while ( $output !== $input );

        return $output;

    }

    /*
     * Focuses on stripping encoded entities
     * *** This appears to be why people use this sample code. Unclear how well Kses does this ***
     *
     * @param   string  $input  Content to be cleaned. It MAY be modified in output
     * @return  string  $input  Modified $input string
     */
    private function strip_encoded_entities( $input ) {

        // Fix &entity\n;
        $input = str_replace(array('&amp;','&lt;','&gt;'), array('&amp;amp;','&amp;lt;','&amp;gt;'), $input);
        $input = preg_replace('/(&#*\w+)[\x00-\x20]+;/u', '$1;', $input);
        $input = preg_replace('/(&#x*[0-9A-F]+);*/iu', '$1;', $input);
        $input = html_entity_decode($input, ENT_COMPAT, 'UTF-8');

        // Remove any attribute starting with "on" or xmlns
        $input = preg_replace('#(<[^>]+?[\x00-\x20"\'])(?:on|xmlns)[^>]*+[>\b]?#iu', '$1>', $input);

        // Remove javascript: and vbscript: protocols
        $input = preg_replace('#([a-z]*)[\x00-\x20]*=[\x00-\x20]*([`\'"]*)[\x00-\x20]*j[\x00-\x20]*a[\x00-\x20]*v[\x00-\x20]*a[\x00-\x20]*s[\x00-\x20]*c[\x00-\x20]*r[\x00-\x20]*i[\x00-\x20]*p[\x00-\x20]*t[\x00-\x20]*:#iu', '$1=$2nojavascript...', $input);
        $input = preg_replace('#([a-z]*)[\x00-\x20]*=([\'"]*)[\x00-\x20]*v[\x00-\x20]*b[\x00-\x20]*s[\x00-\x20]*c[\x00-\x20]*r[\x00-\x20]*i[\x00-\x20]*p[\x00-\x20]*t[\x00-\x20]*:#iu', '$1=$2novbscript...', $input);
        $input = preg_replace('#([a-z]*)[\x00-\x20]*=([\'"]*)[\x00-\x20]*-moz-binding[\x00-\x20]*:#u', '$1=$2nomozbinding...', $input);

        // Only works in IE: <span style="width: expression(alert('Ping!'));"></span>
        $input = preg_replace('#(<[^>]+?)style[\x00-\x20]*=[\x00-\x20]*[`\'"]*.*?expression[\x00-\x20]*\([^>]*+>#i', '$1>', $input);
        $input = preg_replace('#(<[^>]+?)style[\x00-\x20]*=[\x00-\x20]*[`\'"]*.*?behaviour[\x00-\x20]*\([^>]*+>#i', '$1>', $input);
        $input = preg_replace('#(<[^>]+?)style[\x00-\x20]*=[\x00-\x20]*[`\'"]*.*?s[\x00-\x20]*c[\x00-\x20]*r[\x00-\x20]*i[\x00-\x20]*p[\x00-\x20]*t[\x00-\x20]*:*[^>]*+>#iu', '$1>', $input);

        return $input;

    }

    /*
     * Focuses on stripping unencoded HTML tags & namespaces
     *
     * @param   string  $input  Content to be cleaned. It MAY be modified in output
     * @return  string  $input  Modified $input string
     */
    private function strip_tags( $input ) {
        // Remove tags
        $input = preg_replace('#</*(?:applet|b(?:ase|gsound|link)|embed|frame(?:set)?|i(?:frame|layer)|l(?:ayer|ink)|meta|object|s(?:cript|tyle)|title|xml)[^>]*+>#i', '', $input);

        // Remove namespaced elements
        $input = preg_replace('#</*\w+:\w[^>]*+>#i', '', $input);

        return $input;

    }

    /*
     * Focuses on stripping entities from Base64 encoded strings
     *
     * NOT ENABLED by default!
     * To enable 2nd param of clean_input() can be set to anything other than 0 or '0':
     * ie: xssClean->clean_input( $input_string, 1 )
     *
     * @param   string  $input      Maybe Base64 encoded string
     * @return  string  $output     Modified & re-encoded $input string
     */
    private function strip_base64( $input ) {

        $decoded = base64_decode( $input );

        $decoded = $this->strip_tags( $decoded );
        $decoded = $this->strip_encoded_entities( $decoded );

        $output = base64_encode( $decoded );

        return $output;

    }

}

Thanks for it man! it helped :)

Thanks man, help me alot

Thank for your effort

Does this also cover tags with uppercase text. I know that's not proper formatting for a tag, but what if someone is inserting data into their DB after turning it lowercase?

Thanks for your effort:)

Here is the bypass, I think:

<img src =x onerror=confirm(document.cookie);

The regular expression (https://gist.github.com/mbijon/1098477#file-xss_clean-php-L26) expects that attacker will use the closing angular bracket which is missing in the above vector and all browsers will render this ...

Thanks a lot my fellow developer
I think If you take a look at the anti xss library in the link
and compare it with yours you can improve your code even further
https://code.google.com/p/php-antixss/downloads/detail?name=AntiXSS.php&can=2&q=

@soaj1664 voku/anti-xss@bc003bf :)

thanks for a lot
helpful

@ALL -- FYI that I haven't updated this function since 2011 ...so that's likely one of MANY VULNS in this lib. Unless you have performance constraints I recommend HTML Purifier instead of this quick & dirty method.

---

It might be time to update or just pull this lib down.:

• @soaj1664 Thanks for catching that <img src =x onerror=confirm(document.cookie);.
• @voku I'll try running your anit-xss tester from bc003bf against this.

Tks

I tried to add to test DB xss_clean("< a href="#">a</ a>") (without spaces). It added and destroy some data. Is it bug? Or it's not XSS category?

thanks for a lot
helpful

A better alternative to this class https://github.com/ymakux/xss
Ready for production. Based on Kses and Drupal 7 filter

please I need a case that this filter can not catch i try most of cases but these cases were catches please and need at least one case that bypass this filter

rola2010 : test" onmouseover="alert(document.cookie);"

Asynth: unfortunately the code catch this,it seems it can't be broken

I still recommend HTML Purifier or kses instead of this gist.

However:

• UPDATED: This class-based version of xssClean recurses BOTH the encoded entity & tag removal routines. This solves a vulnerability found by 0xmitsurugi & reported privately.
• UPDATED: The exploit reported by @soaj1664 in this comment is fixed. It could have only been effective at the end of the file or if there were no other ">" characters after their exploit ...the problem with filtering that exploint without ONLY a closing ">" is that users would see removal of the entire input body. Instead of removing the entire message body after that exploit, I've chosen to remove up to the next non-word character (such as a newline or file ending char). This could still remove the rest of the input if there are no non-word characters before the end, but it helps keep some of the message in MOST cases, but still making this filter more secure.

👍

thank you so much

Hello, I am new to all xss stuff
how to prevent this

<?php
$_GET['a'] = 'javascript:alert(document.cookie)';
$href = xssClean($_GET['a']);
echo '<a href="'.$href.'">XSS link</a>';
?>

@petrospap

Just echo it with htmlspecialchars(). Example:

$_GET['a'] = 'javascript:alert(document.cookie)';
$href = $_GET['a'];
echo '<a href="'.htmlspecialchars($href).'">XSS link</a>';
?>

pice of code ;)

Hi,

I was testing your filter against a set of XSS test inputs. It seems that your filter is still vulnerable to XSS such as with inputs that contain XSS payloads in the comment-type, anchor and image tags. Examples of one of each are:

<!--#exec cmd=""/usr/X11R6/bin/xterm ?display 127.0.0.1:0 &""-->
<a href="jAvAsCrIpT&colon;alert&lpar;1&rpar;">X</a>
/><img/onerror=\x09javascript:alert(1)\x09src=xxx:x />

A full report can be read in our paper, "Assessment of Dynamic Open-source Cross-site Scripting Filters as Security Devices in Web Applications". I kindly suggest that you add these tags to the blacklist to make it more robust against XSS.

Thank you.

@nat4tq share your paper with us then :)

@nat4tq I tried the code you provided on the local server. This is really a problem. Do you have any solutions?

@nat4tq I tried the code you provided on the local server. This is really a problem. Do you have any solutions?

You could try this library : https://github.com/voku/anti-xss

My method is deprecated. I now recommend http://htmlpurifier.org/