[4.2.x] Fixed CVE-2025-32873 -- Mitigated potential DoS in strip_tags().

sarahboyce · nessita · commit 9cd8028f3e38 · 2025-05-06T22:36:15.000-03:00
Thanks to Elias Myllymäki for the report, and Shai Berger and Jake Howard for the reviews. Co-authored-by: Natalia <124304+nessita@users.noreply.github.com> Backport of 9f3419b from main.
diff --git a/django/utils/html.py b/django/utils/html.py
@@ -17,6 +17,9 @@
 MAX_URL_LENGTH = 2048
 MAX_STRIP_TAGS_DEPTH = 50
 
+# HTML tag that opens but has no closing ">" after 1k+ chars.
+long_open_tag_without_closing_re = _lazy_re_compile(r"<[a-zA-Z][^>]{1000,}")
+
 
 @keep_lazy(SafeString)
 def escape(text):
@@ -175,6 +178,9 @@ def _strip_once(value):
 def strip_tags(value):
     """Return the given HTML with all tags stripped."""
     value = str(value)
+    for long_open_tag in long_open_tag_without_closing_re.finditer(value):
+        if long_open_tag.group().count("<") >= MAX_STRIP_TAGS_DEPTH:
+            raise SuspiciousOperation
     # Note: in typical case this loop executes _strip_once twice (the second
     # execution does not remove any more tags).
     strip_tags_depth = 0
diff --git a/docs/releases/4.2.21.txt b/docs/releases/4.2.21.txt
@@ -7,6 +7,17 @@ Django 4.2.21 release notes
 Django 4.2.21 fixes a security issue with severity "moderate", a data loss bug,
 and a regression in 4.2.20.
 
+CVE-2025-32873: Denial-of-service possibility in ``strip_tags()``
+=================================================================
+
+:func:`~django.utils.html.strip_tags` would be slow to evaluate certain inputs
+containing large sequences of incomplete HTML tags. This function is used to
+implement the :tfilter:`striptags` template filter, which was thus also
+vulnerable.
+
+:func:`~django.utils.html.strip_tags` now raises a :exc:`.SuspiciousOperation`
+exception if it encounters an unusually large number of unclosed opening tags.
+
 Bugfixes
 ========
 
diff --git a/tests/utils_tests/test_html.py b/tests/utils_tests/test_html.py
@@ -115,17 +115,30 @@ def test_strip_tags(self):
             ("><!" + ("&" * 16000) + "D", "><!" + ("&" * 16000) + "D"),
             ("X<<<<br>br>br>br>X", "XX"),
             ("<" * 50 + "a>" * 50, ""),
+            (">" + "<a" * 500 + "a", ">" + "<a" * 500 + "a"),
+            ("<a" * 49 + "a" * 951, "<a" * 49 + "a" * 951),
+            ("<" + "a" * 1_002, "<" + "a" * 1_002),
         )
         for value, output in items:
             with self.subTest(value=value, output=output):
                 self.check_output(strip_tags, value, output)
                 self.check_output(strip_tags, lazystr(value), output)
 
-    def test_strip_tags_suspicious_operation(self):
+    def test_strip_tags_suspicious_operation_max_depth(self):
         value = "<" * 51 + "a>" * 51, "<a>"
         with self.assertRaises(SuspiciousOperation):
             strip_tags(value)
 
+    def test_strip_tags_suspicious_operation_large_open_tags(self):
+        items = [
+            ">" + "<a" * 501,
+            "<a" * 50 + "a" * 950,
+        ]
+        for value in items:
+            with self.subTest(value=value):
+                with self.assertRaises(SuspiciousOperation):
+                    strip_tags(value)
+
     def test_strip_tags_files(self):
         # Test with more lengthy content (also catching performance regressions)
         for filename in ("strip_tags1.html", "strip_tags2.txt"):
