• The CSS URL protocol allowlist is now enforced on the nonstandard -webkit-image-set CSS function. @ltk - #242

• The CSS URL protocol allowlist is now properly enforced in CSS Images Module Level 4 image and image-set functions. @ltk - #240

• Proactively fixed a compatibility issue with libxml >= 2.13.0 (which will be used in an upcoming version of Nokogiri) that caused HTML doctype sanitization to fail. @flavorjones - #238

• Added the text-decoration-skip-ink and text-decoration-thickness CSS properties to the relaxed config. @martineriksson - #228

• CVE-2023-36823: Fixed an HTML CSS sanitization bypass that could allow XSS (cross-site scripting). This issue affects Sanitize versions 3.0.0 through 6.0.1.

  When using Sanitize's relaxed config or a custom config that allows <style> elements and one or more CSS at-rules, carefully crafted input could be used to sneak arbitrary HTML through Sanitize.

  See the following security advisory for additional details: GHSA-f5ww-cq3m-q3g7

  Thanks to @cure53 for finding this issue.

• Sanitize now always removes <noscript> elements and their contents, even when noscript is in the allowlist.

  This fixes a sanitization bypass that could occur when noscript was allowed by a custom allowlist. In this scenario, carefully crafted input could sneak arbitrary HTML through Sanitize, potentially enabling an XSS (cross-site scripting) attack.

  Sanitize's default configs don't allow <noscript> elements and are not vulnerable. This issue only affects users who are using a custom config that adds noscript to the element allowlist.

  The root cause of this issue is that HTML parsing rules treat the contents of a <noscript> element differently depending on whether scripting is enabled in the user agent. Nokogiri doesn't support scripting so it follows the "scripting disabled" rules, but a web browser with scripting enabled will follow the "scripting enabled" rules. This means that Sanitize can't reliably make the contents of a <noscript> element safe for scripting enabled browsers, so the safest thing to do is to remove the element and its contents entirely.

  See the following security advisory for additional details: GHSA-fw3g-2h3j-qmm7

  Thanks to David Klein from TU Braunschweig (@leeN) for reporting this issue.

• Fixed an edge case in which the contents of an "unescaped text" element (such as <noembed> or <xmp>) were not properly escaped if that element was allowlisted and was also inside an allowlisted <math> or <svg> element.

  The only way to encounter this situation was to ignore multiple warnings in the readme and create a custom config that allowlisted all the elements involved, including <math> or <svg>. If you're using a default config or if you heeded the warnings about MathML and SVG not being supported, you're not affected by this issue.

  Please let this be a reminder that Sanitize cannot safely sanitize MathML or SVG content and does not support this use case. The default configs don't allow MathML or SVG elements, and allowlisting MathML or SVG elements in a custom config may create a security vulnerability in your application.

  Documentation has been updated to add more warnings and to make the existing warnings about this more prominent.

  Thanks to David Klein from TU Braunschweig (@leeN) for reporting this issue.

• Ruby 2.5.0 is now the oldest officially supported Ruby version.

• Sanitize now requires Nokogiri 1.12.0 or higher, which includes Nokogumbo. The separate dependency on Nokogumbo has been removed. @lis2 - #211

• Ensure protocol sanitization is applied to data attributes. @ccutrer - #207

• Fixed a deprecation warning in Ruby 2.7 when using keyword arguments in a custom transformer. @mscrivo - #206

• Fixed an HTML sanitization bypass that could allow XSS. This issue affects Sanitize versions 3.0.0 through 5.2.0.

  When HTML was sanitized using the "relaxed" config or a custom config that allows certain elements, some content in a <math> or <svg> element may not have beeen sanitized correctly even if math and svg were not in the allowlist. This could allow carefully crafted input to sneak arbitrary HTML through Sanitize, potentially enabling an XSS (cross-site scripting) attack.

  You are likely to be vulnerable to this issue if you use Sanitize's relaxed config or a custom config that allows one or more of the following HTML elements:

  • iframe
  • math
  • noembed
  • noframes
  • noscript
  • plaintext
  • script
  • style
  • svg
  • xmp

  See the security advisory for more details, including a workaround if you're not able to upgrade: GHSA-p4x4-rw2p-8j8m

  Many thanks to Michał Bentkowski of Securitum for reporting this issue and helping to verify the fix.

• The term "whitelist" has been replaced with "allowlist" throughout Sanitize's source and documentation.

  While the etymology of "whitelist" may not be explicitly racist in origin or intent, there are inherent racial connotations in the implication that white is good and black (as in "blacklist") is not.

  This is a change I should have made long ago, and I apologize for not making it sooner.

• In transformer input, the :is_whitelisted and :node_whitelist keys are now deprecated. New :is_allowlisted and :node_allowlist keys have been added. The old keys will continue to work in order to avoid breaking existing code, but they are no longer documented and may be removed in a future semver major release.

• Added a :parser_options config hash, which makes it possible to pass custom parsing options to Nokogumbo. @austin-wang - #194

• Non-characters and non-whitespace control characters are now stripped from HTML input before parsing to comply with the HTML Standard's preprocessing guidelines. Prior to this Sanitize had adhered to older W3C guidelines that have since been withdrawn. #179

For most users, upgrading from 4.x shouldn't require any changes. However, the minimum required Ruby version has changed, and Sanitize 5.x's HTML output may differ in some small ways from 4.x's output. If this matters to you, please review the changes below carefully.

• Ruby 2.3.0 is now the oldest officially supported Ruby version. Sanitize may work in older 2.x Rubies, but they aren't actively tested. Sanitize definitely no longer works in Ruby 1.9.x.

• Upgraded to Nokogumbo 2.x, which fixes various bugs and adds standard-compliant HTML serialization. @stevecheckoway - #189

• Children of the following elements are now removed by default when these elements are removed, rather than being preserved and escaped:

  • iframe
  • noembed
  • noframes
  • noscript
  • script
  • style

• Children of allowlisted iframe elements are now always removed. In modern HTML, iframe elements should never have children. In HTML 4 and earlier iframe elements were allowed to contain fallback content for legacy browsers, but it's been almost two decades since that was useful.

• Fixed a bug that caused :remove_contents to behave as if it were set to true when it was actually an Array.

• Improved performance and memory usage by optimizing Sanitize#transform_node! @stanhu - #183

• Improved performance slightly by tweaking the order of built-in transformers. @rafbm - #180

• Fixed: A change introduced in 4.6.2 broke certain transformers that relied on being able to mutate the name of an HTML node. That change has been reverted and a test has been added to cover this case. @zetter - #177

• CVE-2018-3740: Fixed an HTML injection vulnerability that could allow XSS.

  When Sanitize <= 4.6.2 is used in combination with libxml2 >= 2.9.2, a specially crafted HTML fragment can cause libxml2 to generate improperly escaped output, allowing non-allowlisted attributes to be used on allowlisted elements.

  Sanitize now performs additional escaping on affected attributes to prevent this.

  Many thanks to the Shopify Application Security Team for responsibly reporting this issue.

• Reduced string allocations to optimize memory usage. @janklimo - #175

• Added support for frozen string literals in Ruby 2.4 . @flavorjones - #174

• Loosened the Nokogumbo dependency to allow installing semver-compatible versions greater than or equal to v1.4. @rafbm - #171

• Added SVG-related CSS properties to the relaxed config. See the diff for the full list of added properties. @louim - #161

• Fixed: Sanitize now strips null bytes (\u0000) before passing input to Nokogumbo, since they can cause recent versions to crash with a failed assertion in the Gumbo parser.

• Added srcset to the attribute allowlist for img elements in the relaxed config. @ejtttje - #156

• Methods can now be used as transformers. @Skipants - #155

• Added -webkit-font-smoothing to the relaxed CSS config. @louim - #154

• Fixed: Nokogumbo >=1.4.9 changed its behavior in a way that allowed invalid doctypes (like <!DOCTYPE nonsense>) when the :allow_doctype config setting was true. Invalid doctypes are now coerced to valid ones as they were prior to this Nokogumbo change.

• Added a new CSS config setting, :import_url_validator. This is a Proc or other callable object that will be called with each @import URL, and should return true to allow the URL or false to remove it. @nikz - #153

• Unpinned the Nokogumbo dependency. @rubys - #141

• Added two new CSS config settings, :at_rules_with_properties and :at_rules_with_styles. These allow you to define which at-rules should be allowed to contain properties and which should be allowed to contain style rules. Previously this was hard-coded internally. #111

  The previous :at_rules setting still exists, and defines at-rules that may not have associated blocks, such as @import. If you have a custom config that contains an :at_rules setting, you may need to move rules can have blocks to either :at_rules_with_properties or :at_rules_with_styles.

  See Sanitize's relaxed config for an example.

• Added full support for CSS @page rules in the relaxed config, including support for all page-margin box rules (such as @top-left, @bottom-center, etc.)

• Added the following CSS at-rules to the relaxed config:

  • @-moz-keyframes
  • @-o-keyframes
  • @-webkit-keyframes
  • @document

• Added a whole bunch of CSS properties to the relaxed config. View the complete list here.

• Small performance improvements.

• Fixed: Upgraded Crass to 1.0.2 to pick up a fix that affected the parsing of CSS @page rules.

• Fixed: Deleting a node in a custom transformer could trigger a memory leak in Nokogiri if that node's children were later reparented, which the built-in CleanElement transformer did by default. The CleanElement transformer is now careful not to reparent the children of deleted nodes. #129

• Fixed: #document and #fragment failed on frozen strings, and could unintentionally modify unfrozen strings if they used an encoding other than UTF-8 or if they contained characters not allowed in HTML. @AnchorCat - #128

• Added the following CSS properties to the relaxed config. @ehudc - #120

  • -moz-text-size-adjust
  • -ms-text-size-adjust
  • -webkit-text-size-adjust
  • text-size-adjust

• Updated Nokogumbo to 1.2.0 to pick up a fix for a Gumbo bug where the entity Æ left its semicolon behind when it was converted to a character during parsing. #119

• Fixed: Harmless whitespace preceding a URL protocol (such as " http://") caused the URL to be removed even when the protocol was allowlisted. @benubois - #126

• Fixed: Some CSS selectors weren't parsed correctly inside the body of a @media block, causing them to be removed even when allowlist rules should have allowed them to remain. #121

• Updated Nokogumbo to 1.1.12, because 1.1.11 silently reverted the change we were trying to pick up in the last release. Now issue #114 is actually fixed.

• Updated Nokogumbo to 1.1.11 to pick up a fix for a Gumbo bug in which certain HTML character entities, such as Ö, were parsed incorrectly, leaving the semicolon behind in the output. #114

As of this version, Sanitize adheres strictly to the SemVer 2.0.0 versioning standard. This release contains API and output changes that are incompatible with previous releases, as indicated by the major version increment.

• HTML is now parsed using Google's Gumbo HTML5 parser, which adheres to the HTML5 parsing spec and behaves much more like modern browser parsers than the previous libxml2-based parser. As a result, HTML output may differ from that of previous versions of Sanitize.

• All transformers now traverse the document from the top down, starting with the first node, then its first child, and so on. The :transformers_breadth config has been removed, and old bottom-up transformers (the previous default) may need to be rewritten.

• Sanitize's built-in configs are now deeply frozen to prevent people from modifying them (either accidentally or maliciously). To customize a built-in config, create a new copy using Sanitize::Config.merge(), like so:

Sanitize.fragment(html, Sanitize::Config.merge(Sanitize::Config::BASIC,
  :elements        => Sanitize::Config::BASIC[:elements]   ['div', 'table'],
  :remove_contents => true
))

• The clean! and clean_document! methods were removed, since they weren't useful and tended to confuse people.

• The clean method was renamed to fragment to more clearly indicate that its intended use is to sanitize an HTML fragment.

• The clean_document method was renamed to document.

• The clean_node! method was renamed to node!.

• The document method now raises a Sanitize::Error if the <html> element isn't allowlisted, rather than a RuntimeError. This error is also now raised regardless of the :remove_contents config setting.

• The :output config has been removed. Output is now always HTML, not XHTML.

• The :output_encoding config has been removed. Output is now always UTF-8.

• Added advanced CSS sanitization support using Crass, which is fully compliant with the CSS Syntax Module Level 3 parsing spec. The contents of allowlisted <style> elements and style attributes in HTML will be sanitized as CSS, or you can use the Sanitize::CSS class to manually sanitize CSS stylesheets or properties.

• Added an :allow_doctype setting. When true, well-formed doctype definitions will be allowed in documents. When false (the default), doctype definitions will be removed from documents. Doctype definitions are never allowed in fragments, regardless of this setting.

• Added the following elements to the relaxed config, in addition to various attributes: article, aside, body, data, div, footer, head, header, html, main, nav, section, span, style, title.

• The :whitespace_elements config is now a Hash, and allows you to specify the text that should be inserted before and after these elements when they're removed. The old-style Array-based config value is still supported for backwards compatibility. @alperkokmen - #94

• Unsuitable Unicode characters are now removed from HTML before it's parsed. #106

• Fixed: Non-tag brackets in input like "1 > 2 and 2 < 1" are now parsed and escaped correctly in accordance with the HTML5 spec, becoming "1 > 2 and 2 < 1". #83

• Fixed: Siblings added after the current node during traversal are now also traversed. In previous versions they were simply skipped. #91

• Fixed: Nokogiri has been smacked and instructed to stop adding newlines after certain elements, because if people wanted newlines there they'd have put them there, dammit. #103

• Fixed: Added a workaround for a libxml2 bug that caused an undesired content-type meta tag to be added to all documents with <head> elements. Nokogiri #1008

• CVE-2018-3740: Fixed an HTML injection vulnerability that could allow XSS (backported from Sanitize 4.6.3). @dometto - #188

  When Sanitize <= 2.1.0 is used in combination with libxml2 >= 2.9.2, a specially crafted HTML fragment can cause libxml2 to generate improperly escaped output, allowing non-allowlisted attributes to be used on allowlisted elements.

  Sanitize now performs additional escaping on affected attributes to prevent this.

  Many thanks to the Shopify Application Security Team for responsibly reporting this issue.

• Added support for allowlisting arbitrary HTML5 data-* attributes. Use the symbol :data instead of an attribute name in the :attributes config to indicate that arbitrary data attributes should be allowed on an element.

• Added the following elements to the relaxed config: address, bdi, hr, and summary.

• Fixed: A colon (:) character in a URL fragment identifier such as #foo:1 was incorrectly treated as a protocol delimiter. @heathd - #87

• Fixed: Version 2.0.5 inadvertently included some work-in-progress changes that shouldn't have made their way into the master branch. This is what happens when I release before coffee instead of after.

• Loosened the Nokogiri dependency back to >= 1.4.4 to allow Sanitize to coexist in newer Rubies with other libraries that restrict Nokogiri to 1.5.x for 1.8.7 compatibility. Sanitize still no longer supports 1.8.7, but this should make life easier for people who need those other libs.

• Added Sanitize.clean_document, which sanitizes a full HTML document rather than just a fragment. [Ben Anderson]

• Nokogiri dependency bumped to 1.6.x.

• Dropped support for Ruby versions older than 1.9.2.

• Loosened the Nokogiri dependency to allow Nokogiri 1.5.x.

• Fixed a bug in which a protocol like "java\script:" would be translated to "java\script:" and allowed through the filter when relative URLs were enabled. This didn't actually allow malicious code to run, but it is undesired behavior.

• Updated the protocol regex to anchor at the beginning of the string rather than the beginning of a line. [Eaden McKee]

• The environment data passed into transformers and the return values expected from transformers have changed. Old transformers will need to be updated. See the README for details.

• Transformers now receive nodes of all types, not just element nodes.

• Sanitize's own core filtering logic is now implemented as a set of always-on transformers.

• The default value for the :output config is now :html. Previously it was :xhtml.

• Added a :whitespace_elements config, which specifies elements (such as <br> and <p>) that should be replaced with whitespace when removed in order to preserve readability. See the README for the default list of elements that will be replaced with whitespace when removed.

• Added a :transformers_breadth config, which may be used to specify transformers that should traverse nodes in a breadth-first mode rather than the default depth-first mode.

• Added the abbr, dfn, kbd, mark, s, samp, time, and var elements to the allowlists for the basic and relaxed configs.

• Added the bdo, del, figcaption, figure, hgroup, ins, rp, rt, ruby, and wbr elements to the allowlist for the relaxed config.

• The dir, lang, and title attributes are now allowlisted for all elements in the relaxed config.

• Bumped minimum Nokogiri version to 1.4.4 to avoid a bug in 1.4.2 (issue #315) that caused </body></html> to be appended to the CDATA inside unterminated script and style elements.

• Added a :remove_contents config setting. If set to true, Sanitize will remove the contents of all non-allowlisted elements in addition to the elements themselves. If set to an array of element names, Sanitize will remove the contents of only those elements (when filtered), and leave the contents of other filtered elements. [Thanks to Rafael Souza for the array option]

• Added an :output_encoding config setting to allow the character encoding for HTML output to be specified. The default is utf-8.

• The environment hash passed into transformers now includes a :node_name item containing the lowercase name of the current HTML node (e.g. "div").

• Returning anything other than a Hash or nil from a transformer will now raise a meaningful Sanitize::Error exception rather than an unintended NameError.

• Requires Nokogiri ~> 1.4.1.

• Added support for transformers, which allow you to filter and alter nodes using your own custom logic, on top of (or instead of) Sanitize's core filter. See the README for details and examples.

• Added Sanitize.clean_node!, which sanitizes a Nokogiri::XML::Node and all its children.

• Added elements <h1> through <h6> to the Relaxed allowlist. [Suggested by David Reese]

• Migrated from Hpricot to Nokogiri. Requires libxml2 >= 2.7.2 [Adam Hooper]

• Added an :output config setting to allow the output format to be specified. Supported formats are :xhtml (the default) and :html (which outputs HTML4).

• Changed protocol regex to ensure Sanitize doesn't kill URLs with colons in path segments. [Peter Cooper]

• Added a workaround for an Hpricot bug that prevents attribute names from being downcased in recent versions of Hpricot. This was exploitable to prevent non-allowlisted protocols from being cleaned. [Reported by Ben Wanicur]

• Requires Hpricot 0.8.1 , which is finally compatible with Ruby 1.9.1.

• Fixed a bug that caused named character entities containing digits (like ²) to be escaped when they shouldn't have been. [Reported by Sebastian Steinmetz]

• Removed htmlentities gem dependency.

• Existing well-formed character entity references in the input string are now preserved rather than being decoded and re-encoded.

• The ' character is now encoded as ' instead of ' to prevent problems in IE6.

• You can now specify the symbol :all in place of an element name in the attributes config hash to allow certain attributes on all elements. [Thanks to Mutwin Kraus]

• Fixed a bug introduced in version 1.0.3 that prevented non-allowlisted protocols from being cleaned when relative URLs were allowed. [Reported by Dev Purkayastha]

• Fixed "undefined method `parent='" exceptions caused by parser changes in edge Hpricot.

• Fixed a bug that made it possible to sneak a non-allowlisted element through by repeating it several times in a row. All versions of Sanitize prior to 1.0.4 are vulnerable. [Reported by Cristobal]

• Fixed a bug whereby incomplete Unicode or hex entities could be used to prevent non-allowlisted protocols from being cleaned. Since IE6 and Opera still decode the incomplete entities, users of those browsers may be vulnerable to malicious script injection on websites using versions of Sanitize prior to 1.0.3.

• Fixed a bug that caused an exception to be thrown when parsing a valueless attribute that's expected to contain a URL.

• You can now specify :relative in a protocol config array to allow attributes containing relative URLs with no protocol. The Basic and Relaxed configs have been updated to allow relative URLs.

• Added a workaround for an Hpricot bug that causes HTML entities for non-ASCII characters to be replaced by question marks, and all other entities to be destructively decoded.

• First release.