=begin Copyright 2010-2022 Ecsypno This file is part of the Arachni Framework project and is subject to redistribution and commercial restrictions. Please see the Arachni Framework web site for more information on licensing and terms of use. =end # Injects JS taint code and checks to see if it gets executed as proof of # vulnerability. # # @author Tasos "Zapotek" Laskos # # @see http://cwe.mitre.org/data/definitions/79.html # @see http://ha.ckers.org/xss.html # @see http://secunia.com/advisories/9716/ class Arachni::Checks::XssScriptContext < Arachni::Check::Base ATTRIBUTES = [ 'onload', 'onunload', 'onblur', 'onchange', 'onfocus', 'onreset', 'onselect', 'onsubmit', 'onabort', 'onkeydown', 'onkeypress', 'onkeyup', 'onclick', 'ondblclick', 'onmousedown', 'onmousemove', 'onmouseout', 'onmouseover', 'onmouseup', # Not an event attribute so it gets special treatment by being checked # for a "script:" prefix. 'src' ] class SAX attr_reader :tainted def initialize( seed ) @seed = seed @attributes = Set.new( ATTRIBUTES ) end def document end def tainted? !!@tainted end def start_element( name ) @in_script = (name.to_s.downcase == 'script') end def end_element( name ) @in_script = false end def attr( name, value ) name = name.to_s.downcase value = value.downcase return if !@attributes.include?( name ) if name == 'src' if @seed.start_with?( 'javascript:' ) && value == @seed @tainted = true fail Arachni::Parser::SAX::Stop end else if value == @seed @tainted = true fail Arachni::Parser::SAX::Stop end end if value.include?( @seed ) @tainted = true fail Arachni::Parser::SAX::Stop end end def text( value ) return if !@in_script || value !~ /#{Regexp.escape( @seed )}/i @tainted = true fail Arachni::Parser::SAX::Stop end end def self.seed 'window.top._%s_taint_tracer.log_execution_flow_sink()' end def self.strings return @strings if @strings @strings ||= [ "javascript:#{seed}" ] ['\'', '"', ''].each do |quote| [ "%q;#{seed}%q", "%q;#{seed};%q" ].each do |payload| @strings << payload.gsub( '%q', quote ) end end [ "1;#{seed}%q", "1;\n#{seed}%q" ].each do |payload| ['', ';'].each do |s| @strings << payload.gsub( '%q', s ) end end @strings = @strings.map { |s| [ s, "#{s}//" ] }.flatten @strings << "*/;\n#{seed}/*" # In case they're placed as assoc array values. @strings << seed @strings << "\",x:#{seed},y:\"" @strings << "',x:#{seed},y:'" @strings << "" end def self.options @options ||= { format: [ Format::STRAIGHT ] } end def self.optimization_cache @optimization_cache ||= {} end def optimization_cache self.class.optimization_cache end def taints( browser_cluster ) self.class.strings.map { |taint| taint % browser_cluster.javascript_token } end def run with_browser_cluster do |cluster| audit taints( cluster ), self.class.options do |response, element| next if !response.html? # Completely body based, identical bodies will yield identical # results. k = "#{response.url.hash}-#{response.body.hash}".hash next if optimization_cache[k] optimization_cache[k] = true check_and_log( response, element ) end end end def check_and_log( response, element ) # Check to see if the response is tainted before going any further, # this also serves as a rudimentary check for really simple cases. return if !(proof = tainted?( response, element.seed )) if proof.is_a? String log vector: element, proof: element.seed, response: response return end with_browser_cluster do |cluster| print_info 'Response is tainted, scheduling a taint-trace.' # Pass the response to the BrowserCluster for evaluation and see if the # JS payload we injected got executed by inspecting the page's # execution-flow sink. cluster.trace_taint( response, { taint: self.class.seed, args: [element, page] }, self.class.check_browser_result_cb ) end end def self.check_browser_result( result, element, referring_page, cluster ) page = result.page print_info 'Checking results of deferred taint analysis for' << ' execution-flow sink data.' return if page.dom.execution_flow_sinks.empty? log( vector: element, proof: element.seed, page: page, referring_page: referring_page ) end def self.check_browser_result_cb @check_browser_result_cb ||= method(:check_browser_result) end def tainted?( response, seed ) return if seed.to_s.empty? || !response.body.to_s.include?( seed ) handler = SAX.new( self.class.seed % browser_cluster.javascript_token ) Arachni::Parser.parse( response.body, handler: handler ) handler.tainted? end def self.info { name: 'XSS in script context', description: %q{ Injects JS taint code and check to see if it gets executed as proof of vulnerability. }, elements: [ Element::Form, Element::Link, Element::Cookie, Element::NestedCookie, Element::Header, Element::LinkTemplate ], author: 'Tasos "Zapotek" Laskos ', version: '0.2.6', issue: { name: %q{Cross-Site Scripting (XSS) in script context}, description: %q{ Client-side scripts are used extensively by modern web applications. They perform from simple functions (such as the formatting of text) up to full manipulation of client-side data and Operating System interaction. Cross Site Scripting (XSS) allows clients to inject scripts into a request and have the server return the script to the client in the response. This occurs because the application is taking untrusted data (in this example, from the client) and reusing it without performing any validation or sanitisation. If the injected script is returned immediately this is known as reflected XSS. If the injected script is stored by the server and returned to any client visiting the affected page, then this is known as persistent XSS (also stored XSS). Arachni has discovered that it is possible to force the page to execute custom JavaScript code. }, references: { 'Secunia' => 'http://secunia.com/advisories/9716/', 'WASC' => 'http://projects.webappsec.org/w/page/13246920/Cross%20Site%20Scripting', 'OWASP' => 'https://www.owasp.org/index.php/XSS_%28Cross_Site_Scripting%29_Prevention_Cheat_Sheet' }, tags: %w(xss script dom injection), cwe: 79, severity: Severity::HIGH, remedy_guidance: %q{ To remedy XSS vulnerabilities, it is important to never use untrusted or unfiltered data within the code of a HTML page. Untrusted data can originate not only form the client but potentially a third party or previously uploaded file etc. Filtering of untrusted data typically involves converting special characters to their HTML entity encoded counterparts (however, other methods do exist, see references). These special characters include: * `&` * `<` * `>` * `"` * `'` * `/` An example of HTML entity encoding is converting `<` to `<`. Although it is possible to filter untrusted input, there are five locations within an HTML page where untrusted input (even if it has been filtered) should never be placed: 1. Directly in a script. 2. Inside an HTML comment. 3. In an attribute name. 4. In a tag name. 5. Directly in CSS. Each of these locations have their own form of escaping and filtering. _Because many browsers attempt to implement XSS protection, any manual verification of this finding should be conducted using multiple different browsers and browser versions._ } } } end end