Grav is Vulnerable to XXE via SVG Upload

Overview

Dear Grav Security Team,

A security vulnerability was discovered in Grav CMS that allows authenticated attackers to read arbitrary files from the server through XML External Entity (XXE) injection.

Vulnerability Summary

Technical Details

Root Cause The application uses simplexml_load_string() to process uploaded SVG files without disabling external entity loading. This allows attackers to inject XXE payloads that are processed by the XML parser.

Vulnerable Code Pattern

// Current (Vulnerable):
$svg = simplexml_load_string($content);

// No LIBXML_NOENT flag or entity loader protection

Attack Vector

1. Attacker authenticates to Grav admin panel
2. Uploads malicious SVG file via Pages → Media or File Manager plugin
3. Server parses SVG and processes XXE entities
4. Arbitrary file contents are exfiltrated

Impact

An authenticated attacker can:

1. Read sensitive files:

   • /etc/passwd - System user information
   • user/accounts/*.yaml - Admin credentials and 2FA secrets
   • user/config/system.yaml - System configuration
   • .env files - Environment secrets and API keys

2. Perform SSRF - Access internal services via external entity URLs

3. Potential DoS - Billion laughs attack via recursive entity expansion

Proof of Concept

Malicious SVG Payload

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE svg [
  <!ENTITY xxe SYSTEM "file:///etc/passwd">
]>
<svg xmlns="http://www.w3.org/2000/svg" width="100" height="100">
  <text x="10" y="50">&xxe;</text>
</svg>

Steps to Reproduce

1. Login to Grav CMS admin panel
2. Navigate to Pages → select any page → Media tab
3. Upload the malicious SVG file
4. Observe file contents in response/error or stored output

Recommended Fix

Option 1: Add XXE Protection Flags

libxml_use_internal_errors(true);
$svg = simplexml_load_string($content, 'SimpleXMLElement', LIBXML_NOENT | LIBXML_DTDLOAD);

Option 2: Use SVG Sanitizer Library (Recommended)

use enshrined\svgSanitize\Sanitizer;

$sanitizer = new Sanitizer();
$sanitizer->removeRemoteReferences(true);
$cleanSVG = $sanitizer->sanitize($content);

The enshrined/svg-sanitize library properly strips XXE payloads and other malicious SVG content.

Request

1. Please acknowledge receipt of this report within 5 business days
2. Please provide an estimated timeline for a security patch
3. I am happy to assist with testing the fix
4. I request a CVE be assigned for this vulnerability
5. If you have a security advisory process, please include me in the credits

Turki Almatrafi.

---

Maintainer note — fix applied (2026-04-24)

Fixed across two repos:

1. Grav core on the 2.0 branch (commit 5a12f9be8, ships in 2.0.0-beta.2) — VectorImageMedium::__construct (the code path that reads width/height from an uploaded SVG) now strips <!DOCTYPE> and <!ENTITY> declarations before parsing, and calls simplexml_load_string with LIBXML_NONET | LIBXML_NOERROR | LIBXML_NOWARNING. On PHP < 8 it also calls libxml_disable_entity_loader(true) for the duration of the parse.

2. rhukster/dom-sanitizer (commit 02d08ec) — the library Grav ships as its SVG sanitizer. loadDocument now applies the same DOCTYPE/ENTITY strip and passes LIBXML_NONET to loadXML/loadHTML.

With both layers in place, the PoC:

<!DOCTYPE svg [
  <!ENTITY xxe SYSTEM "file:///etc/passwd">
]>
<svg xmlns="http://www.w3.org/2000/svg" width="100" height="100">
  <text x="10" y="50">&xxe;</text>
</svg>

no longer expands &xxe;, and the parser cannot make outbound filesystem or network requests for external entities/DTDs. Billion-laughs-style entity expansion is also neutralized because the declarations are stripped before libxml ever sees them.

Files:

• system/src/Grav/Common/Page/Medium/VectorImageMedium.php.
• tests/unit/Grav/Common/Security/SvgXxeSecurityTest.php — XXE neutralization + billion-laughs + plain-SVG regression.
• dom-sanitizer: src/DOMSanitizer.php + two new XXE tests in its own suite.

Affected packages

• getgrav/grav < 2.0.0-beta.2

Remediation

Upgrade to a patched release:

• getgrav/grav 2.0.0-beta.2