Validateur SVG

// Parses the SVG and lists what a browser could execute. Nothing here
// inserts the markup into the page, which is what keeps it safe to run
// against a file you do not trust.
const EXECUTABLE = ['javascript:', 'vbscript:', 'data:text/html'];

function auditSvg(source) {
  const doc = new DOMParser().parseFromString(source, 'image/svg+xml');
  if (doc.querySelector('parsererror')) {
    return { wellFormed: false, findings: [] };
  }

  const findings = [];
  const walk = (node) => {
    const tag = node.localName.toLowerCase();
    if (tag === 'script' || tag === 'handler' || tag === 'foreignobject') {
      findings.push(node.nodeName + ' element');
    }
    for (const attr of node.attributes) {
      const value = attr.value.trim().toLowerCase();
      if (attr.name.toLowerCase().startsWith('on')) {
        findings.push(attr.name + ' event handler');
      }
      if (EXECUTABLE.some((scheme) => value.startsWith(scheme))) {
        findings.push('executable URI in ' + attr.name);
      }
    }
    for (const child of node.children) walk(child);
  };

  walk(doc.documentElement);
  return { wellFormed: true, findings };
}

# resolve_entities and no_network are the flags that matter: an SVG with a
# DOCTYPE is an XXE vector like any other XML document.
from lxml import etree

DANGEROUS = {'script', 'handler', 'foreignObject'}
EXECUTABLE = ('javascript:', 'vbscript:', 'data:text/html')

parser = etree.XMLParser(
    resolve_entities=False,
    no_network=True,
    load_dtd=False,
    huge_tree=False,
)
root = etree.fromstring(svg_bytes, parser)   # XMLSyntaxError if malformed

for el in root.iter():
    if not isinstance(el.tag, str):
        continue                             # comment or processing instruction
    if etree.QName(el).localname in DANGEROUS:
        print(el.sourceline, etree.QName(el).localname, 'element')
    for name, value in el.attrib.items():
        local = etree.QName(name).localname if name.startswith('{') else name
        if local.lower().startswith('on'):
            print(el.sourceline, name, 'event handler')
        if value.strip().lower().startswith(EXECUTABLE):
            print(el.sourceline, name, 'executable URI')

import javax.xml.XMLConstants;
import javax.xml.parsers.SAXParserFactory;
import org.xml.sax.Attributes;
import org.xml.sax.InputSource;
import org.xml.sax.helpers.DefaultHandler;
import java.util.Locale;
import java.util.Set;

SAXParserFactory factory = SAXParserFactory.newInstance();
factory.setNamespaceAware(true);
factory.setFeature(XMLConstants.FEATURE_SECURE_PROCESSING, true);
factory.setFeature("http://xml.org/sax/features/external-general-entities", false);
factory.setFeature("http://xml.org/sax/features/external-parameter-entities", false);
// Old Illustrator exports carry a DOCTYPE. Drop the next line only if you
// must accept them, and keep the two features above false either way.
factory.setFeature("http://apache.org/xml/features/disallow-doctype-decl", true);

final Set<String> dangerous = Set.of("script", "handler", "foreignObject");

factory.newSAXParser().parse(
    new InputSource(new java.io.StringReader(svg)),
    new DefaultHandler() {
        @Override
        public void startElement(String uri, String local, String qName, Attributes atts) {
            if (dangerous.contains(local)) {
                System.out.println("element: " + qName);
            }
            for (int i = 0; i < atts.getLength(); i++) {
                String name = atts.getQName(i);
                String value = atts.getValue(i).trim().toLowerCase(Locale.ROOT);
                if (name.toLowerCase(Locale.ROOT).startsWith("on")) {
                    System.out.println("event handler: " + name);
                }
                if (value.startsWith("javascript:") || value.startsWith("data:text/html")) {
                    System.out.println("executable URI in " + name);
                }
            }
        }
    });

using System;
using System.Xml;

var dangerous = new[] { "script", "handler", "foreignObject" };
var settings = new XmlReaderSettings
{
    DtdProcessing = DtdProcessing.Prohibit,  // no DOCTYPE, no entity expansion
    XmlResolver = null,                      // never fetch anything
    MaxCharactersFromEntities = 1024 * 1024,
};

using var reader = XmlReader.Create(new StringReader(svg), settings);
var lineInfo = (IXmlLineInfo)reader;

while (reader.Read())
{
    if (reader.NodeType != XmlNodeType.Element) continue;

    if (Array.IndexOf(dangerous, reader.LocalName) >= 0)
        Console.WriteLine($"line {lineInfo.LineNumber}: <{reader.Name}>");

    while (reader.MoveToNextAttribute())
    {
        var value = reader.Value.Trim().ToLowerInvariant();
        if (reader.LocalName.StartsWith("on", StringComparison.OrdinalIgnoreCase))
            Console.WriteLine($"line {lineInfo.LineNumber}: {reader.Name} handler");
        if (value.StartsWith("javascript:") || value.StartsWith("data:text/html"))
            Console.WriteLine($"line {lineInfo.LineNumber}: URI in {reader.Name}");
    }
    reader.MoveToElement();
}

<?php
// LIBXML_NONET blocks every external fetch, including a DOCTYPE's DTD.
libxml_use_internal_errors(true);

$doc = new DOMDocument();
if (!$doc->loadXML($svg, LIBXML_NONET)) {
    fwrite(STDERR, "not well-formed XML\n");
    exit(1);
}

$xpath = new DOMXPath($doc);
$xpath->registerNamespace('svg', 'http://www.w3.org/2000/svg');

foreach ($xpath->query('//svg:script | //svg:handler | //svg:foreignObject') as $node) {
    printf("%s element on line %d\n", $node->nodeName, $node->getLineNo());
}

foreach ($xpath->query('//@*') as $attr) {
    $name = strtolower($attr->nodeName);
    $value = strtolower(trim($attr->nodeValue));
    if (str_starts_with($name, 'on') || str_starts_with($value, 'javascript:')) {
        printf("%s on line %d\n", $attr->nodeName, $attr->getLineNo());
    }
}
// To clean rather than report, use enshrined/svg-sanitize.

# Well-formedness first. An SVG that fails here fails as XML.
xmllint --noout --nonet icon.svg

# A smoke test for executable content. This is a grep, not a sanitiser: it
# misses base64 payloads, <set attributeName="onload">, and obfuscation.
grep -niE 'script|foreignObject|on[a-z]+ *=|javascript:|data:text/html' icon.svg

# Check a directory of exports and keep going after each failure:
find . -name '*.svg' -print0 | xargs -0 -n1 xmllint --noout --nonet

# svgo is a minifier and does not remove script by default.
# For sanitising, use DOMPurify in Node with its SVG profile.