Cleaner: Remove SVG image data URLs since they can embed script content.

Reported as GHSL-2021-1038

2 files changed, 60 insertions, 8 deletions

diff --git a/src/lxml/html/clean.py b/src/lxml/html/clean.py
index 25844e87..dd3a28ad 100644
--- a/src/lxml/html/clean.py
+++ b/src/lxml/html/clean.py
@@ -75,18 +75,25 @@ _looks_like_tag_content = re.compile(
 
 # All kinds of schemes besides just javascript: that can cause
 # execution:
-_is_image_dataurl = re.compile(
-    r'^data:image/.+;base64', re.I).search
+_find_image_dataurls = re.compile(
+    r'^data:image/(.+);base64,', re.I).findall
 _is_possibly_malicious_scheme = re.compile(
-    r'(?:javascript|jscript|livescript|vbscript|data|about|mocha):',
-    re.I).search
+    r'(javascript|jscript|livescript|vbscript|data|about|mocha):',
+    re.I).findall
+# SVG images can contain script content
+_is_unsafe_image_type = re.compile(r"(xml|svg)", re.I).findall
+
 def _is_javascript_scheme(s):
-    if _is_image_dataurl(s):
-        return None
-    return _is_possibly_malicious_scheme(s)
+    is_image_url = False
+    for image_type in _find_image_dataurls(s):
+        is_image_url = True
+        if _is_unsafe_image_type(image_type):
+            return True
+    if is_image_url:
+        return False
+    return bool(_is_possibly_malicious_scheme(s))
 
 _substitute_whitespace = re.compile(r'[\s\x00-\x08\x0B\x0C\x0E-\x19]+').sub
-# FIXME: should data: be blocked?
 
 # FIXME: check against: http://msdn2.microsoft.com/en-us/library/ms537512.aspx
 _conditional_comment_re = re.compile(
diff --git a/src/lxml/html/tests/test_clean.py b/src/lxml/html/tests/test_clean.py
index d395d514..a05d9673 100644
--- a/src/lxml/html/tests/test_clean.py
+++ b/src/lxml/html/tests/test_clean.py
@@ -1,3 +1,5 @@
+import base64
+import gzip
 import unittest
 from lxml.tests.common_imports import make_doctest
 
@@ -143,6 +145,49 @@ class CleanerTest(unittest.TestCase):
                 cleaned,
                 "%s  ->  %s" % (style_code, cleaned))
 
+    def test_svg_data_links(self):
+        # Remove SVG images with potentially insecure content.
+        svg = b'<svg onload="alert(123)" />'
+        svgz = gzip.compress(svg)
+        svg_b64 = base64.b64encode(svg).decode('ASCII')
+        svgz_b64 = base64.b64encode(svgz).decode('ASCII')
+        urls = [
+            "data:image/svg+xml;base64," + svg_b64,
+            "data:image/svg+xml-compressed;base64," + svgz_b64,
+        ]
+        for url in urls:
+            html = '<img src="%s">' % url
+            s = lxml.html.fragment_fromstring(html)
+
+            cleaned = lxml.html.tostring(clean_html(s))
+            self.assertEqual(
+                b'<img src="">',
+                cleaned,
+                "%s  ->  %s" % (url, cleaned))
+
+    def test_image_data_links(self):
+        data = b'123'
+        data_b64 = base64.b64encode(data).decode('ASCII')
+        urls = [
+            "data:image/jpeg;base64," + data_b64,
+            "data:image/apng;base64," + data_b64,
+            "data:image/png;base64," + data_b64,
+            "data:image/gif;base64," + data_b64,
+            "data:image/webp;base64," + data_b64,
+            "data:image/bmp;base64," + data_b64,
+            "data:image/tiff;base64," + data_b64,
+            "data:image/x-icon;base64," + data_b64,
+        ]
+        for url in urls:
+            html = '<img src="%s">' % url
+            s = lxml.html.fragment_fromstring(html)
+
+            cleaned = lxml.html.tostring(clean_html(s))
+            self.assertEqual(
+                html.encode("UTF-8"),
+                cleaned,
+                "%s  ->  %s" % (url, cleaned))
+
     def test_formaction_attribute_in_button_input(self):
         # The formaction attribute overrides the form's action and should be
         # treated as a malicious link attribute