loofah 2.2.0 → 2.2.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 6656f9e5edc815b2c5ee676d1c4fb818b2dc03f4
-  data.tar.gz: 7bea1d04f8af479fd825c7adf687f0ca0c624830
+  metadata.gz: 55986c6edc7d96bf6301f01ed6c8474a0e0eedc3
+  data.tar.gz: e1e5ef44e194bd9333cfc92f9d182bedaac5799c
 SHA512:
-  metadata.gz: 42f030b7228867ebf322c9d8e286349e1288ef3d60f90fe404b0d9250cc626ea6fad84ff1325cd2754ea4a7fdf80802a4bdae5a9b7121ac312e56d96c280d1a3
-  data.tar.gz: 8a67c56281a65b6e89d8623f40423ae41ed2628eeb0a90193196cfb87aeb4efccbe23c961b05ab26a247bac0117a55b68dea97ab6b67076e272ebad8471e33cb
+  metadata.gz: 7d45dbc3cb1e275e81f127e92107ba540da3f8294d486e927325cd9a8cb8c742ba1f34c14101b46437242b3778911c3f7392bf822b2dbeccd835410f52ca757f
+  data.tar.gz: 70318ae204e34718bbdf818c6be1d302e5377eec163ff1380556e219443d1d0c42887f5e6e69b4d2ef8876e6cbbaa4f2f7f84d73fe71e13bbe1b2d10b199e072

data/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## 2.2.1 / 2018-03-19
+
+Addresses CVE-2018-8048. Loofah allowed non-whitelisted attributes to be present in sanitized output when input with specially-crafted HTML fragments.
+
+This CVE's public notice is at https://github.com/flavorjones/loofah/issues/144
+
+
 ## 2.2.0 / 2018-02-11
 
 Features:

data/Manifest.txt

@@ -5,6 +5,7 @@ MIT-LICENSE.txt
 Manifest.txt
 README.md
 Rakefile
+SECURITY.md
 benchmark/benchmark.rb
 benchmark/fragment.html
 benchmark/helper.rb
@@ -14,6 +15,7 @@ lib/loofah/elements.rb
 lib/loofah/helpers.rb
 lib/loofah/html/document.rb
 lib/loofah/html/document_fragment.rb
+lib/loofah/html5/libxml2_workarounds.rb
 lib/loofah/html5/scrub.rb
 lib/loofah/html5/whitelist.rb
 lib/loofah/instance_methods.rb

data/README.md

@@ -310,14 +310,22 @@ And the IRC channel is \#loofah on freenode.
 
 ## Security
 
-Some tools may incorrectly report loofah is a potential security
-vulnerability. Loofah depends on Nokogiri, and it's possible to use
-Nokogiri in a dangerous way (by enabling its DTDLOAD option and
-disabling its NONET option).  This dangerous Nokogiri configuration,
-which is sometimes used by other components, can create an XML
-External Entity (XXE) vulnerability if the XML data is not trusted.
-However, loofah never enables this dangerous Nokogiri configuration;
-loofah never enables DTDLOAD, and it never disables NONET.
+See [`SECURITY.md`](SECURITY.md) for vulnerability reporting details.
+
+
+### "Secure by Default"
+
+Some tools may incorrectly report Loofah as a potential security
+vulnerability.
+
+Loofah depends on Nokogiri, and it's _possible_ to use Nokogiri in a
+dangerous way (by enabling its DTDLOAD option and disabling its NONET
+option). This specifically allows the opportunity for an XML External
+Entity (XXE) vulnerability if the XML data is untrusted.
+
+However, Loofah __never enables this Nokogiri configuration__; Loofah
+never enables DTDLOAD, and it never disables NONET, thereby protecting
+you by default from this XXE vulnerability.
 
 
 ## Related Links
@@ -345,7 +353,7 @@ And a big shout-out to Corey Innis for the name, and feedback on the API.
 
 ## Thank You
 
-The following people have generously donated via the Pledgie[http://pledgie.com] badge on the {Loofah github page}[https://github.com/flavorjones/loofah]:
+The following people have generously donated via the [Pledgie](http://pledgie.com) badge on the [Loofah github page](https://github.com/flavorjones/loofah):
 
 * Bill Harding
 

data/Rakefile

@@ -12,9 +12,9 @@ Hoe.spec "loofah" do
   developer "Mike Dalessio", "mike.dalessio@gmail.com"
   developer "Bryan Helmkamp", "bryan@brynary.com"
 
-  self.extra_rdoc_files = FileList["*.rdoc"]
+  self.extra_rdoc_files = FileList["*.md"]
   self.history_file     = "CHANGELOG.md"
-  self.readme_file      = "README.rdoc"
+  self.readme_file      = "README.md"
   self.license          "MIT"
 
   extra_deps     << ["nokogiri", ">=1.5.9"]

data/SECURITY.md

@@ -0,0 +1,18 @@
+# Security and Vulnerability Reporting
+
+The Loofah core contributors take security very seriously and investigate all reported vulnerabilities.
+
+If you would like to report a vulnerablity or have a security concern regarding Loofah, please [report it via HackerOne](https://hackerone.com/loofah/reports/new).
+
+Your report will be acknowledged within 24 hours, and you'll receive a more detailed response within 72 hours indicating next steps in handling your report.
+
+If you have not received a reply to your submission within 48 hours, there are a few steps you can take:
+
+* Contact the current security coordinator (Mike Dalessio <mike.dalessio@gmail.com>)
+* Email the Loofah user group at loofah@librelist.com (archive at http://librelist.com)
+
+Please note, the user group list is a public area. When escalating in that venue, please do not discuss your issue. Simply say that you're trying to get a hold of someone from the core team.
+
+The information you share with the Loofah core contributors as part of this process will be kept confidential within the team, unless or until we need to share information upstream with our dependent libraries' core teams, at which point we will notify you.
+
+If a vulnerability is first reported by you, we will credit you with the discovery in the public disclosure.

data/lib/loofah.rb

@@ -6,6 +6,7 @@ require 'loofah/metahelpers'
 require 'loofah/elements'
 
 require 'loofah/html5/whitelist'
+require 'loofah/html5/libxml2_workarounds'
 require 'loofah/html5/scrub'
 
 require 'loofah/scrubber'
@@ -27,7 +28,7 @@ require 'loofah/html/document_fragment'
 #
 module Loofah
   # The version of Loofah you are using
-  VERSION = '2.2.0'
+  VERSION = '2.2.1'
 
   class << self
     # Shortcut for Loofah::HTML::Document.parse

data/lib/loofah/html5/libxml2_workarounds.rb

@@ -0,0 +1,26 @@
+# coding: utf-8
+require 'set'
+
+module Loofah
+  #
+  #  constants related to working around unhelpful libxml2 behavior
+  #
+  #  ಠ_ಠ
+  #
+  module LibxmlWorkarounds
+    #
+    #  these attributes and qualifying parent tags are determined by the code at:
+    #
+    #    https://git.gnome.org/browse/libxml2/tree/HTMLtree.c?h=v2.9.2#n714
+    #
+    #  see comments about CVE-2018-8048 within the tests for more information
+    #
+    BROKEN_ESCAPING_ATTRIBUTES = Set.new %w[
+        href
+        action
+        src
+        name
+      ]
+    BROKEN_ESCAPING_ATTRIBUTES_QUALIFYING_TAG = {"name" => "a"}
+  end
+end

data/lib/loofah/html5/scrub.rb

@@ -1,5 +1,3 @@
-#encoding: US-ASCII
-
 require 'cgi'
 require 'crass'
 
@@ -65,6 +63,8 @@ module Loofah
           node.attribute_nodes.each do |attr_node|
             node.remove_attribute(attr_node.name) if attr_node.value !~ /[^[:space:]]/
           end
+
+          force_correct_attribute_escaping! node
         end
 
         def scrub_css_attribute node
@@ -100,6 +100,35 @@ module Loofah
 
           Crass::Parser.stringify sanitized_tree
         end
+
+        private
+
+        #
+        #  libxml2 >= 2.9.2 fails to escape comments within some attributes.
+        #
+        #  see comments about CVE-2018-8048 within the tests for more information
+        #
+        def force_correct_attribute_escaping! node
+          return unless Nokogiri::VersionInfo.instance.libxml2?
+
+          node.attribute_nodes.each do |attr_node|
+            next unless LibxmlWorkarounds::BROKEN_ESCAPING_ATTRIBUTES.include?(attr_node.name)
+
+            tag_name = LibxmlWorkarounds::BROKEN_ESCAPING_ATTRIBUTES_QUALIFYING_TAG[attr_node.name]
+            next unless tag_name.nil? || tag_name == node.name
+
+            #
+            #  this block is just like CGI.escape in Ruby 2.4, but
+            #  only encodes space and double-quote, to mimic
+            #  pre-2.9.2 behavior
+            #
+            encoding = attr_node.value.encoding
+            attr_node.value = attr_node.value.gsub(/[ "]/) do |m|
+              '%' + m.unpack('H2' * m.bytesize).join('%').upcase
+            end.force_encoding(encoding)
+          end
+        end
+
       end
     end
   end

data/test/integration/test_ad_hoc.rb

@@ -188,6 +188,71 @@ mso-bidi-language:#0400;}
       html = "<p>Foo</p>\n<p>Bar</p>"
       assert_equal "Foo\nBar", Loofah.scrub_document(html, :prune).text
     end
+
+    #
+    #  tests for CVE-2018-8048 (see https://github.com/flavorjones/loofah/issues/144)
+    #
+    #  libxml2 >= 2.9.2 fails to escape comments within some attributes. It
+    #  wants to ensure these comments can be treated as "server-side includes",
+    #  but as a result fails to ensure that serialization is well-formed,
+    #  resulting in an opportunity for XSS injection of code into a final
+    #  re-parsed document (presumably in a browser).
+    #
+    #  we'll test this by parsing the HTML, serializing it, then
+    #  re-parsing it to ensure there isn't any ambiguity in the output
+    #  that might allow code injection into a browser consuming
+    #  "sanitized" output.
+    #
+    [
+      #
+      #  these tags and attributes are determined by the code at:
+      #
+      #    https://git.gnome.org/browse/libxml2/tree/HTMLtree.c?h=v2.9.2#n714
+      #
+      {tag: "a",   attr: "href"},
+      {tag: "div", attr: "href"},
+      {tag: "a",   attr: "action"},
+      {tag: "div", attr: "action"},
+      {tag: "a",   attr: "src"},
+      {tag: "div", attr: "src"},
+      {tag: "a",   attr: "name"},
+      #
+      #  note that div+name is _not_ affected by the libxml2 issue.
+      #  but we test it anyway to ensure our logic isn't modifying
+      #  attributes that don't need modifying.
+      #
+      {tag: "div", attr: "name", unescaped: true},
+    ].each do |config|
+
+      define_method "test_uri_escaping_of_#{config[:attr]}_attr_in_#{config[:tag]}_tag" do
+        html = %{<#{config[:tag]} #{config[:attr]}='examp<!--" unsafeattr=foo()>-->le.com'>test</#{config[:tag]}>}
+
+        reparsed = Loofah.fragment(Loofah.fragment(html).scrub!(:prune).to_html)
+        attributes = reparsed.at_css(config[:tag]).attribute_nodes
+
+        assert_equal [config[:attr]], attributes.collect(&:name)
+        if Nokogiri::VersionInfo.new.libxml2?
+          if config[:unescaped]
+            #
+            #  this attribute was emitted wrapped in single-quotes, so a double quote is A-OK.
+            #  assert that this attribute's serialization is unaffected.
+            #
+            assert_equal %{examp<!--" unsafeattr=foo()>-->le.com}, attributes.first.value
+          else
+            #
+            #  let's match the behavior in libxml < 2.9.2.
+            #  test that this attribute's serialization is well-formed and sanitized.
+            #
+            assert_equal %{examp<!--%22%20unsafeattr=foo()>-->le.com}, attributes.first.value
+          end
+        else
+          #
+          #  yay for consistency in javaland. move along, nothing to see here.
+          #
+          assert_equal %{examp<!--%22 unsafeattr=foo()>-->le.com}, attributes.first.value
+        end
+      end
+
+    end
   end
 end
-

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: loofah
 version: !ruby/object:Gem::Version
-  version: 2.2.0
+  version: 2.2.1
 platform: ruby
 authors:
 - Mike Dalessio
@@ -9,7 +9,7 @@ authors:
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2018-02-11 00:00:00.000000000 Z
+date: 2018-03-19 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: nokogiri
@@ -193,7 +193,18 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: '3.16'
-description: ''
+description: |-
+  Loofah is a general library for manipulating and transforming HTML/XML
+  documents and fragments. It's built on top of Nokogiri and libxml2, so
+  it's fast and has a nice API.
+
+  Loofah excels at HTML sanitization (XSS prevention). It includes some
+  nice HTML sanitizers, which are based on HTML5lib's whitelist, so it
+  most likely won't make your codes less secure. (These statements have
+  not been evaluated by Netexperts.)
+
+  ActiveRecord extensions for sanitization are available in the
+  [`loofah-activerecord` gem](https://github.com/flavorjones/loofah-activerecord).
 email:
 - mike.dalessio@gmail.com
 - bryan@brynary.com
@@ -204,6 +215,7 @@ extra_rdoc_files:
 - MIT-LICENSE.txt
 - Manifest.txt
 - README.md
+- SECURITY.md
 files:
 - ".gemtest"
 - CHANGELOG.md
@@ -212,6 +224,7 @@ files:
 - Manifest.txt
 - README.md
 - Rakefile
+- SECURITY.md
 - benchmark/benchmark.rb
 - benchmark/fragment.html
 - benchmark/helper.rb
@@ -221,6 +234,7 @@ files:
 - lib/loofah/helpers.rb
 - lib/loofah/html/document.rb
 - lib/loofah/html/document_fragment.rb
+- lib/loofah/html5/libxml2_workarounds.rb
 - lib/loofah/html5/scrub.rb
 - lib/loofah/html5/whitelist.rb
 - lib/loofah/instance_methods.rb
@@ -242,14 +256,14 @@ files:
 - test/unit/test_helpers.rb
 - test/unit/test_scrubber.rb
 - test/unit/test_scrubbers.rb
-homepage: 
+homepage: https://github.com/flavorjones/loofah
 licenses:
 - MIT
 metadata: {}
 post_install_message: 
 rdoc_options:
 - "--main"
-- README.rdoc
+- README.md
 require_paths:
 - lib
 required_ruby_version: !ruby/object:Gem::Requirement
@@ -267,5 +281,6 @@ rubyforge_project:
 rubygems_version: 2.6.12
 signing_key: 
 specification_version: 4
-summary: ''
+summary: Loofah is a general library for manipulating and transforming HTML/XML documents
+  and fragments
 test_files: []