[1.4.x] Made is_safe_url() reject URLs that start with control characters.

This is a security fix; disclosure to follow shortly.

3 files changed, 30 insertions, 2 deletions

diff --git a/django/utils/http.py b/django/utils/http.py
index e69a92b578..b8c81a8418 100644
--- a/django/utils/http.py
+++ b/django/utils/http.py
@@ -4,6 +4,7 @@ import re
 import sys
 import urllib
 import urlparse
+import unicodedata
 from email.utils import formatdate
 
 from django.utils.datastructures import MultiValueDict
@@ -232,9 +233,10 @@ def is_safe_url(url, host=None):
 
     Always returns ``False`` on an empty url.
     """
+    if url is not None:
+        url = url.strip()
     if not url:
         return False
-    url = url.strip()
     # Chrome treats \ completely as /
     url = url.replace('\\', '/')
     # Chrome considers any URL with more than two slashes to be absolute, but
@@ -248,5 +250,10 @@ def is_safe_url(url, host=None):
     # allow this syntax.
     if not url_info[1] and url_info[0]:
         return False
+    # Forbid URLs that start with control characters. Some browsers (like
+    # Chrome) ignore quite a few control characters at the start of a
+    # URL and might consider the URL as scheme relative.
+    if unicodedata.category(unicode(url[0]))[0] == 'C':
+        return False
     return (not url_info[1] or url_info[1] == host) and \
         (not url_info[0] or url_info[0] in ['http', 'https'])
diff --git a/docs/releases/1.4.20.txt b/docs/releases/1.4.20.txt
index 9c46c5510f..f2ca5ac103 100644
--- a/docs/releases/1.4.20.txt
+++ b/docs/releases/1.4.20.txt
@@ -5,3 +5,22 @@ Django 1.4.20 release notes
 *March 18, 2015*
 
 Django 1.4.20 fixes one security issue in 1.4.19.
+
+Mitigated possible XSS attack via user-supplied redirect URLs
+=============================================================
+
+Django relies on user input in some cases (e.g.
+:func:`django.contrib.auth.views.login` and :doc:`i18n </topics/i18n/index>`)
+to redirect the user to an "on success" URL. The security checks for these
+redirects (namely ``django.utils.http.is_safe_url()``) accepted URLs with
+leading control characters and so considered URLs like ``\x08javascript:...``
+safe. This issue doesn't affect Django currently, since we only put this URL
+into the ``Location`` response header and browsers seem to ignore JavaScript
+there. Browsers we tested also treat URLs prefixed with control characters such
+as ``%08//example.com`` as relative paths so redirection to an unsafe target
+isn't a problem either.
+
+However, if a developer relies on ``is_safe_url()`` to
+provide safe redirect targets and puts such a URL into a link, they could
+suffer from an XSS attack as some browsers such as Google Chrome ignore control
+characters at the start of a URL in an anchor ``href``.
diff --git a/tests/regressiontests/utils/http.py b/tests/regressiontests/utils/http.py
index 3ec237ae64..8245a7e54e 100644
--- a/tests/regressiontests/utils/http.py
+++ b/tests/regressiontests/utils/http.py
@@ -98,7 +98,9 @@ class TestUtilsHttp(unittest.TestCase):
                         'http:\/example.com',
                         'http:/\example.com',
                         'javascript:alert("XSS")'
-                        '\njavascript:alert(x)'):
+                        '\njavascript:alert(x)',
+                        '\x08//example.com',
+                        '\n'):
             self.assertFalse(http.is_safe_url(bad_url, host='testserver'), "%s should be blocked" % bad_url)
         for good_url in ('/view/?param=http://example.com',
                      '/view/?param=https://example.com',