Loofah is a general library for manipulating and transforming HTML/XML documents and fragments. It's built on top of Nokogiri and libxml2, so it's fast and has a nice API.
Loofah excels at HTML sanitization (XSS prevention). It includes some nice HTML sanitizers, which are based on HTML5lib's whitelist, so it most likely won't make your codes less secure. (These statements have not been evaluated by Netexperts.)
ActiveRecord extensions for sanitization are available in the
- Easily write custom scrubbers for HTML/XML leveraging the sweetness of Nokogiri (and HTML5lib's whitelists).
- Common HTML sanitizing tasks are built-in:
- Strip unsafe tags, leaving behind only the inner text.
- Prune unsafe tags and their subtrees, removing all traces that they ever existed.
- Escape unsafe tags and their subtrees, leaving behind lots of < and > entities.
- Whitewash the markup, removing all attributes and namespaced nodes.
- Common HTML transformation tasks are built-in:
- Add the nofollow attribute to all hyperlinks.
- Format markup as plain text, with or without sensible whitespace handling around block elements.
- Replace Rails's
sanitizeview helper methods.
Compare and Contrast
Loofah is one of two known Ruby XSS/sanitization solutions that guarantees well-formed and valid markup (the other is Sanitize, which also uses Nokogiri).
Loofah works on XML, XHTML and HTML documents.
Also, it's pretty fast. Here is a benchmark comparing Loofah to other commonly-used libraries (ActionView, Sanitize, HTML5lib and HTMLfilter):
Lastly, Loofah is extensible. It's super-easy to write your own custom scrubbers for whatever document manipulation you need. You don't like the built-in scrubbers? Build your own, like a boss.
Loofah wraps Nokogiri in a loving embrace. Nokogiri is an excellent HTML/XML parser. If you don't know how Nokogiri works, you might want to pause for a moment and go check it out. I'll wait.
Loofah presents the following classes:
The documents and fragments are subclasses of the similar Nokogiri classes.
The Scrubber represents the document manipulation, either by wrapping a block,
span2div = ::. do |node| node.name = "div" if node.name == "span" end
or by implementing a method.
Side Note: Fragments vs Documents
Generally speaking, unless you expect to have a DOCTYPE and a single
root node, you don't have a document, you have a fragment. For
HTML, another rule of thumb is that documents have
tags, and fragments usually do not.
HTML fragments should be parsed with Loofah.fragment. The result won't
be wrapped in
body tags, won't have a DOCTYPE declaration,
head elements will be silently ignored, and multiple root nodes are
XML fragments should be parsed with Loofah.xml_fragment. The result won't have a DOCTYPE declaration, and multiple root nodes are allowed.
HTML documents should be parsed with Loofah.document. The result will
have a DOCTYPE declaration, along with
XML documents should be parsed with Loofah.xml_document. The result will have a DOCTYPE declaration and a single root node.
Loofah::HTML::Document and Loofah::HTML::DocumentFragment
These classes are subclasses of Nokogiri::HTML::Document and Nokogiri::HTML::DocumentFragment, so you get all the markup fixer-uppery and API goodness of Nokogiri.
The module methods Loofah.document and Loofah.fragment will parse an HTML document and an HTML fragment, respectively.
.(unsafe_html).is_a?(Nokogiri::HTML::Document) # => true .(unsafe_html).is_a?(Nokogiri::HTML::DocumentFragment) # => true
Loofah injects a
scrub! method, which takes either a symbol (for
built-in scrubbers) or a Loofah::Scrubber object (for custom
scrubbers), and modifies the document in-place.
to_s to return HTML:
unsafe_html = "ohai! <div>div is safe</div> <script>but script is not</script>" doc = .(unsafe_html).scrub!(:prune) doc.to_s # => "ohai! <div>div is safe</div> "
text to return plain text:
doc.text # => "ohai! div is safe "
to_text is available, which does the right thing with
whitespace around block-level elements.
doc = .("<h1>Title</h1><div>Content</div>") doc.text # => "TitleContent" # probably not what you want doc.to_text # => "\nTitle\n\nContent\n" # better
Loofah::XML::Document and Loofah::XML::DocumentFragment
These classes are subclasses of Nokogiri::XML::Document and Nokogiri::XML::DocumentFragment, so you get all the markup fixer-uppery and API goodness of Nokogiri.
The module methods Loofah.xml_document and Loofah.xml_fragment will parse an XML document and an XML fragment, respectively.
.(bad_xml).is_a?(Nokogiri::XML::Document) # => true .(bad_xml).is_a?(Nokogiri::XML::DocumentFragment) # => true
Nodes and NodeSets
Nokogiri::XML::Node and Nokogiri::XML::NodeSet also get a
method, which makes it easy to scrub subtrees.
The following code will apply the
employee_scrubber only to the
employee nodes (and their subtrees) in the document:
And this code will only scrub the first
employee node and its subtree:
A Scrubber wraps up a block (or method) that is run on a document node:
# change all <span> tags to <div> tags span2div = ::. do |node| node.name = "div" if node.name == "span" end
This can then be run on a document:
.("<span>foo</span><p>bar</p>").scrub!(span2div).to_s # => "<div>foo</div><p>bar</p>"
Scrubbers can be run on a document in either a top-down traversal (the default) or bottom-up. Top-down scrubbers can optionally return Scrubber::STOP to terminate the traversal of a subtree. Read below and in the Loofah::Scrubber class for more detailed usage.
Here's an XML example:
# remove all <employee> tags that have a "deceased" attribute set to true bring_out_your_dead = ::. do |node| if node.name == "employee" and node["deceased"] == "true" node.remove :::: # don't bother with the rest of the subtree end end .(File.read('plague.xml')).scrub!(bring_out_your_dead)
=== Built-In HTML Scrubbers
Loofah comes with a set of sanitizing scrubbers that use HTML5lib's whitelist algorithm:
doc.scrub!(:strip) # replaces unknown/unsafe tags with their inner text doc.scrub!(:prune) # removes unknown/unsafe tags and their children doc.scrub!(:escape) # escapes unknown/unsafe tags, like this: <script> doc.scrub!(:whitewash) # removes unknown/unsafe/namespaced tags and their children, # and strips all node attributes
Loofah also comes with some common transformation tasks:
doc.scrub!(:nofollow) # adds rel="nofollow" attribute to links doc.scrub!(:unprintable) # removes unprintable characters from text nodes
See Loofah::Scrubbers for more details and example usage.
You can chain scrubbers:
.("<span>hello</span> <script>alert('OHAI')</script>") \ .scrub!(:prune) \ .scrub!(span2div).to_s # => "<div>hello</div> "
The class methods Loofah.scrub_fragment and Loofah.scrub_document are shorthand.
.(unsafe_html, :prune) .(unsafe_html, :prune) .(bad_xml, custom_scrubber) .(bad_xml, custom_scrubber)
are the same thing as (and arguably semantically clearer than):
.(unsafe_html).scrub!(:prune) .(unsafe_html).scrub!(:prune) .(bad_xml).scrub!(custom_scrubber) .(bad_xml).scrub!(custom_scrubber)
Loofah has two "view helpers": Loofah::Helpers.sanitize and
Loofah::Helpers.strip_tags, both of which are drop-in replacements for
the Rails ActionView helpers of the same name.
These are no longer required automatically. You must require
- Nokogiri >= 1.5.9
- gem install loofah
The bug tracker is available here:
And the mailing list is on librelist:
And the IRC channel is #loofah on freenode.
SECURITY.md for vulnerability reporting details.
"Secure by Default"
Some tools may incorrectly report Loofah as a potential security vulnerability.
Loofah depends on Nokogiri, and it's possible to use Nokogiri in a dangerous way (by enabling its DTDLOAD option and disabling its NONET option). This specifically allows the opportunity for an XML External Entity (XXE) vulnerability if the XML data is untrusted.
However, Loofah never enables this Nokogiri configuration; Loofah never enables DTDLOAD, and it never disables NONET, thereby protecting you by default from this XXE vulnerability.
- Nokogiri: http://nokogiri.org
- libxml2: http://xmlsoft.org
- html5lib: https://code.google.com/p/html5lib
Featuring code contributed by:
- Aaron Patterson
- John Barnette
- Josh Owens
- Paul Dix
- Luke Melia
And a big shout-out to Corey Innis for the name, and feedback on the API.
- Bill Harding
This library was formerly known as Dryopteris, which was a very bad name that nobody could spell properly.
Distributed under the MIT License. See
MIT-LICENSE.txt for details.