SafeHTML is anti-XSS HTML parser, written in PHP.

written and distributed under the BSD License

About this project

This parser strips down all potentially dangerous content within HTML:

  • opening tag without its closing tag
  • closing tag without its opening tag
  • any of these tags: "base", "basefont", "head", "html", "body", "applet", "object",
    "iframe", "frame", "frameset", "script", "layer", "ilayer", "embed", "bgsound",
    "link", "meta", "style", "title", "blink", "xml" etc.
  • any of these attributes: on*, data*, dynsrc
  • javascript:/vbscript:/about: etc. protocols
  • expression/behavior etc. in styles
  • any other active content

It also tries to convert code to valid HTML, but htmltidy is far better solution for this task.

Whats new?


  • added HTML5 tag terminators for paragraph
  • removed obsolete and deprecated HTML elements


  • added new HTML5 Block-level elements


  • Replaced preg_replace() e modifier with preg_replace_callback


  • UTF-7 XSS vulnerability fixed


  • Allowed tags with setAllowTags() method.
  • AllowTags can be disabled using resetAllowTags()


  • Added 'dl' to the list of 'lists' tags.
  • Added 'callto' to the white list of protocols.
  • Added white list of "namespaced" attributes.


  • More accurate UTF-7 decoding.


  • Two serious security flaws fixed: UTF-7 XSS and CSS comments handling.


  • Security flaw (improper quotes handling in attributes' values) fixed. Big thanks to Nick Cleaton.


  • Dumb bug fixed (some closing tags were ignored).


  • Two holes (with decimal HTML entities and with \x00 symbol) fixed.
  • Class rewritten under PEAR coding standards.
  • Class now uses unmodified HTMLSax3 from PEAR.
  • To the list of table tags added: "caption", "col", "colgroup".


  • It was possible to create XSS with hexadecimal HTML entities. Fixed. Big thanks to Christian Stocker.


  • "id" and "name" attributes added to dangerous attributes list, because malefactor can broke legal javascript by spoofing ID or NAME of some element.
  • New method parse() allows to do all parsing process in two lines of code. Examples also updated.
  • New array, closeParagraph, contains list of block-level elements. When we open such element, we should close paragraph before. . It allows SafeHTML to produce more XHTML compliant code.
  • Added "webcal" to white list of protocols for those who uses calendar programs (Mozilla/iCal/etc).
  • Now SafeHTML strips down table elements when we are not inside table.
  • Now SafeHTML correctly closes unclosed "li" tags: before opening "li" of the same nesting level.


  • New "dangerous" protocols: hcp, ms-help, help, disk,, opera, res, resource, chrome, mocha, livescript.
  • <XML> tag was moved from "tags for deletion" to "tags for deletion with content".
  • New "dangerous" CSS instruction "include-source" (NN4 specific).
  • New array, Attributes, contains list of attributes for removal. If you need to remove "id" or "name" attribute, just add it to this array.
  • Now it is possible to choose between white-list and black-list filtering of protocols. Defaults are "white-list".
This list is: "http", "https", "ftp", "telnet", "news", "nntp", "gopher", "mailto", "file".
  • For speed purposes, we now filter protocols only from these attributes: src, href, action, lowsrc, dynsrc, background, codebase.
  • Opera6 XSS bug ([\xC0][\xBC]script>alert(1)[\xC0][\xBC]/script> [UTF-8] workarounded.


New "dangerous" tag: plaintext.


Added array of elements that can have no closing tag.


Bug fix: <img src="javascript:alert(1);"> attack.
Thanks to shmel.


Bug fix: safehtml hangs on <style></style></style> code.
Thanks to lj user=electrocat.


First public release

Advanced Topics

Dependencies for SafeHTML

  • PHP 8.0
  • HTMLSax3 3.0.0

Authors and copyrights

Author: Roman Ivanov.
© 2006–2023, WackoWiki Team
© 2004–2005, Roman Ivanov
© 2004–2005, Pixel-Apes
© 2004–2005, JetStyle


If you found any bugs in this parser, please inform us — Bugtracker.

Show Files (2 files)

Read comments (2 comments)