rehype-sanitize

What is rehype-sanitize?

The rehype-sanitize package is a plugin for the rehype ecosystem that allows you to sanitize HTML. It is used to clean up and secure HTML content by removing potentially harmful elements and attributes, making it safe to use in web applications.

What are rehype-sanitize's main functionalities?

Basic Sanitization

This feature allows you to sanitize basic HTML content by removing potentially harmful elements like <script> tags. The example code demonstrates how to use rehype-sanitize to clean up an HTML string.

const rehype = require('rehype');
const rehypeSanitize = require('rehype-sanitize');

const html = '<div><script>alert("XSS")</script><p>Hello World</p></div>';

rehype()
  .use(rehypeSanitize)
  .process(html)
  .then((file) => {
    console.log(String(file));
  });

Custom Schema

This feature allows you to use a custom schema for sanitization. The example code demonstrates how to use a GitHub-flavored schema to sanitize HTML content.

const rehype = require('rehype');
const rehypeSanitize = require('rehype-sanitize');
const schema = require('hast-util-sanitize/lib/github');

const html = '<div><script>alert("XSS")</script><p>Hello World</p></div>';

rehype()
  .use(rehypeSanitize, schema)
  .process(html)
  .then((file) => {
    console.log(String(file));
  });

Customizing Allowed Elements

This feature allows you to customize the allowed elements and attributes in the HTML content. The example code demonstrates how to allow only <p>, <b>, and <i> tags and the 'className' attribute.

const rehype = require('rehype');
const rehypeSanitize = require('rehype-sanitize');

const schema = {
  tagNames: ['p', 'b', 'i'],
  attributes: {
    '*': ['className']
  }
};

const html = '<div class="container"><p class="text">Hello <b>World</b></p></div>';

rehype()
  .use(rehypeSanitize, schema)
  .process(html)
  .then((file) => {
    console.log(String(file));
  });

Other packages similar to rehype-sanitize

sanitize-html

sanitize-html is a library for cleaning up user-submitted HTML, preserving whitelisted elements and attributes while removing potentially harmful content. It offers a high degree of customization and is widely used for sanitizing HTML in web applications. Compared to rehype-sanitize, sanitize-html provides more granular control over the sanitization process and is not tied to the rehype ecosystem.

dompurify

DOMPurify is a fast, tolerant, and secure client-side HTML, MathML, and SVG sanitizer. It works in all modern browsers and can be used to clean up HTML content before rendering it in the DOM. DOMPurify is known for its performance and security features, making it a popular choice for client-side sanitization. Unlike rehype-sanitize, which is a plugin for the rehype ecosystem, DOMPurify is a standalone library.

xss

xss is a module used to filter input from users to prevent XSS attacks. It provides a set of default rules for sanitizing HTML and allows for customization of these rules. The xss library is designed to be simple and effective, making it a good choice for server-side sanitization. Compared to rehype-sanitize, xss is more focused on preventing XSS attacks and is not part of the rehype ecosystem.

README

rehype-sanitize

rehype plugin to sanitize HTML.

Contents

• What is this?
• When should I use this?
• Install
• Use
• API
  • unified().use(rehypeSanitize[, schema])
• Example
  • Example: headings (DOM clobbering)
  • Example: math
  • Example: syntax highlighting
• Types
• Compatibility
• Security
• Related
• Contribute
• License

What is this?

This package is a unified (rehype) plugin to make sure HTML is safe. It drops anything that isn’t explicitly allowed by a schema (defaulting to how github.com works).

unified is a project that transforms content with abstract syntax trees (ASTs). rehype adds support for HTML to unified. hast is the HTML AST that rehype uses. This is a rehype plugin that transforms hast.

When should I use this?

It’s recommended to sanitize your HTML any time you do not completely trust authors or the plugins being used.

This plugin is built on hast-util-sanitize, which cleans hast syntax trees. rehype focusses on making it easier to transform content by abstracting such internals away.

Install

This package is ESM only. In Node.js (version 12.20+, 14.14+, or 16.0+), install with npm:

npm install rehype-sanitize

In Deno with Skypack:

import rehypeSanitize from 'https://cdn.skypack.dev/rehype-sanitize@5?dts'

In browsers with Skypack:

<script type="module">
  import rehypeSanitize from 'https://cdn.skypack.dev/rehype-sanitize@5?min'
</script>

Use

Say we have the following file index.html:

<div onmouseover="alert('alpha')">
  <a href="jAva script:alert('bravo')">delta</a>
  <img src="x" onerror="alert('charlie')">
  <iframe src="javascript:alert('delta')"></iframe>
  <math>
    <mi xlink:href="data:x,<script>alert('echo')</script>"></mi>
  </math>
</div>
<script>
require('child_process').spawn('echo', ['hack!']);
</script>

And our module example.js looks as follows:

import {read} from 'to-vfile'
import {unified} from 'unified'
import rehypeParse from 'rehype-parse'
import rehypeSanitize from 'rehype-sanitize'
import rehypeStringify from 'rehype-stringify'

main()

async function main() {
  const file = await unified()
    .use(rehypeParse, {fragment: true})
    .use(rehypeSanitize)
    .use(rehypeStringify)
    .process(await read('index.html'))

  console.log(String(file))
}

Now running node example.js yields:

<div>
  <a>delta</a>
  <img src="x">




</div>

API

This package exports the following identifiers: defaultSchema. The default export is rehypeSanitize.

unified().use(rehypeSanitize[, schema])

Sanitize HTML.

schema

Sanitation schema that defines if and how nodes and properties should be cleaned. The default schema is exported as defaultSchema.

This option is a bit advanced as it requires knowledge of ASTs, so we defer to the documentation available for Schema in hast-util-sanitize.

Example

Example: headings (DOM clobbering)

DOM clobbering is an attack in which malicious HTML confuses an application by naming elements, through id or name attributes, such that they overshadow presumed properties in window (the global scope in browsers). DOM clobbering often occurs when user content is used to generate heading IDs. To illustrate, say we have this browser.js file:

console.log(current)

And our module example.js contains:

import {promises as fs} from 'node:fs'
import {unified} from 'unified'
import rehypeParse from 'rehype-parse'
import rehypeStringify from 'rehype-stringify'

main()

async function main() {
  const browser = String(await fs.readFile('browser.js'))
  const document = `<a name="old"></a>
<h1 id="current">Current</h1>
${`<p>${'Lorem ipsum dolor sit amet. '.repeat(20)}</p>\n`.repeat(20)}
<p>Link to <a href="#current">current</a>, link to <a href="#old">old</a>.`

  const file = await unified()
    .use(rehypeParse, {fragment: true})
    .use(() => (tree) => {
      tree.children.push({
        type: 'element',
        tagName: 'script',
        properties: {type: 'module'},
        children: [{type: 'text', value: browser}]
      })
    })
    .use(rehypeStringify)
    .process(document)

  await fs.writeFile('output.html', file.value)
}

This code processes HTML, inlines our browser script into it, and writes it out. The input HTML models how markdown often looks on platforms like GitHub, which allow heading IDs to be generated from their text and embedded HTML (including <a name="old"></a>, which can be used to create anchors for renamed headings to prevent links from breaking). The generated HTML looks like:

<a name="old"></a>
<h1 id="current">Current</h1>
<p>Lorem ipsum dolor sit amet.<!--…--></p>
<p>Link to <a href="#current">current</a>, link to <a href="#old">old</a>.</p>
<script type="module">console.log(current)</script>

When you run this code locally, open the generated output.html, you can observe that the links at the bottom work, but also that the <h1> element is printed to the console (the clobbering). rehype-sanitize solved the clobbering by prefixing every id and name attribute with 'user-content-'. Changing example.js:

@@ -15,6 +15,7 @@ ${`<p>${'Lorem ipsum dolor sit amet. '.repeat(20)}</p>\n`.repeat(20)}

   const file = await unified()
     .use(rehypeParse, {fragment: true})
+    .use(rehypeSanitize)
     .use(() => (tree) => {
       tree.children.push({
         type: 'element',

Now yields:

-<a name="old"></a>
-<h1 id="current">Current</h1>
+<a name="user-content-old"></a>
+<h1 id="user-content-current">Current</h1>

But this introduces another problem as the links are now broken. It could perhaps be solved by changing all links, but that would make the links rather ugly, and we’d need to track what IDs we have outside of the user content on our pages too. Alternatively, and what arguably looks better, we could rewrite pretty links to their safe but ugly prefixed elements. This is what GitHub does. Replace browser.js with the following:

// Page load (you could wrap this in a DOM ready if the script is loaded early).
hashchange()

// When URL changes.
window.addEventListener('hashchange', hashchange)

// When on the URL already, perhaps after scrolling, and clicking again, which
// doesn’t emit `hashchange`.
document.addEventListener(
  'click',
  (event) => {
    if (
      event.target &&
      event.target instanceof HTMLAnchorElement &&
      event.target.href === location.href &&
      location.hash.length > 1
    ) {
      setTimeout(() => {
        if (!event.defaultPrevented) {
          hashchange()
        }
      })
    }
  },
  false
)

function hashchange() {
  /** @type {string|undefined} */
  let hash

  try {
    hash = decodeURIComponent(location.hash.slice(1)).toLowerCase()
  } catch {
    return
  }

  const name = 'user-content-' + hash
  const target =
    document.getElementById(name) || document.getElementsByName(name)[0]

  if (target) {
    setTimeout(() => {
      target.scrollIntoView()
    }, 0)
  }
}

Example: math

Math can be enabled in rehype by using the plugins rehype-katex or rehype-mathjax. The operate on spans and divs with certain classes and inject complex markup and of inline styles, most of which this plugin will remove. Say our module example.js contains:

import {unified} from 'unified'
import rehypeParse from 'rehype-parse'
import rehypeKatex from 'rehype-katex'
import rehypeSanitize from 'rehype-sanitize'
import rehypeStringify from 'rehype-stringify'

main()

async function main() {
  const file = await unified()
    .use(rehypeParse, {fragment: true})
    .use(rehypeKatex)
    .use(rehypeSanitize)
    .use(rehypeStringify)
    .process('<span class="math math-inline">L</span>')

  console.log(String(file))
}

Running that yields:

<span><span><span>LL</span><span aria-hidden="true"><span><span></span><span>L</span></span></span></span></span>

It is possible to pass a schema which allows MathML and inline styles, but it would be complex, and allows all inline styles, which is unsafe. Alternatively, and arguably better, would be to first sanitize the HTML, allowing only the specific classes that rehype-katex and rehype-mathjax use, and then using those plugins:

@@ -1,7 +1,7 @@
 import {unified} from 'unified'
 import rehypeParse from 'rehype-parse'
 import rehypeKatex from 'rehype-katex'
-import rehypeSanitize from 'rehype-sanitize'
+import rehypeSanitize, {defaultSchema} from 'rehype-sanitize'
 import rehypeStringify from 'rehype-stringify'

 main()
@@ -9,8 +9,21 @@ main()
 async function main() {
   const file = await unified()
     .use(rehypeParse, {fragment: true})
+    .use(rehypeSanitize, {
+      ...defaultSchema,
+      attributes: {
+        ...defaultSchema.attributes,
+        div: [
+          ...(defaultSchema.attributes.div || []),
+          ['className', 'math', 'math-display']
+        ],
+        span: [
+          ...(defaultSchema.attributes.span || []),
+          ['className', 'math', 'math-inline']
+        ]
+      }
+    })
     .use(rehypeKatex)
-    .use(rehypeSanitize)
     .use(rehypeStringify)
     .process('<span class="math math-inline">L</span>')

Example: syntax highlighting

Highlighting, for example with rehype-highlight, can be solved similar to how math is solved (see previous example). That is, use rehype-sanitize and allow the classes needed for highlighting, and highlight afterwards:

import {unified} from 'unified'
import rehypeParse from 'rehype-parse'
import rehypeHighlight from 'rehype-highlight'
import rehypeSanitize, {defaultSchema} from './index.js'
import rehypeStringify from 'rehype-stringify'

main()

async function main() {
  const file = await unified()
    .use(rehypeParse, {fragment: true})
    .use(rehypeSanitize, {
      ...defaultSchema,
      attributes: {
        ...defaultSchema.attributes,
        code: [
          ...(defaultSchema.attributes.code || []),
          // List of all allowed languages:
          ['className', 'language-js', 'language-css', 'language-md']
        ]
      }
    })
    .use(rehypeHighlight, {subset: false})
    .use(rehypeStringify)
    .process('<pre><code className="language-js">console.log(1)</code></pre>')

  console.log(String(file))
}

Alternatively, it’s possible to make highlighting safe by allowing all the classes used on tokens. Modifying the above code like so:

 async function main() {
   const file = await unified()
     .use(rehypeParse, {fragment: true})
+    .use(rehypeHighlight, {subset: false})
     .use(rehypeSanitize, {
       ...defaultSchema,
       attributes: {
         ...defaultSchema.attributes,
-        code: [
-          ...(defaultSchema.attributes.code || []),
-          // List of all allowed languages:
-          ['className', 'hljs', 'language-js', 'language-css', 'language-md']
+        span: [
+          ...(defaultSchema.attributes.span || []),
+          // List of all allowed tokens:
+          ['className', 'hljs-addition', 'hljs-attr', 'hljs-attribute', 'hljs-built_in', 'hljs-bullet', 'hljs-char', 'hljs-code', 'hljs-comment', 'hljs-deletion', 'hljs-doctag', 'hljs-emphasis', 'hljs-formula', 'hljs-keyword', 'hljs-link', 'hljs-literal', 'hljs-meta', 'hljs-name', 'hljs-number', 'hljs-operator', 'hljs-params', 'hljs-property', 'hljs-punctuation', 'hljs-quote', 'hljs-regexp', 'hljs-section', 'hljs-selector-attr', 'hljs-selector-class', 'hljs-selector-id', 'hljs-selector-pseudo', 'hljs-selector-tag', 'hljs-string', 'hljs-strong', 'hljs-subst', 'hljs-symbol', 'hljs-tag', 'hljs-template-tag', 'hljs-template-variable', 'hljs-title', 'hljs-type', 'hljs-variable'
+          ]
         ]
       }
     })
-    .use(rehypeHighlight, {subset: false})
     .use(rehypeStringify)
     .process('<pre><code className="language-js">console.log(1)</code></pre>')

Types

This package is fully typed with TypeScript. It exports an Options type, which specifies the interface of the accepted options.

Compatibility

Projects maintained by the unified collective are compatible with all maintained versions of Node.js. As of now, that is Node.js 12.20+, 14.14+, and 16.0+. Our projects sometimes work with older versions, but this is not guaranteed.

This plugin works with rehype-parse version 3+, rehype-stringify version 3+, rehype version 5+, and unified version 6+.

Security

The defaults are safe but improper use of rehype-sanitize can open you up to a cross-site scripting (XSS) attack.

Use rehype-sanitize after the last unsafe thing: everything after rehype-sanitize could be unsafe (but is fine if you do trust it).

Related

• hast-util-sanitize — utility to sanitize hast
• rehype-format — format HTML
• rehype-minify — minify HTML

Contribute

See contributing.md in rehypejs/.github for ways to get started. See support.md for ways to get help.

This project has a code of conduct. By interacting with this repository, organization, or community you agree to abide by its terms.

License

MIT © Titus Wormer