openstack
/
bandit
Code Issues Proposed changes
bandit/doc/old/xml.md

7.6 KiB
Raw Blame History

Use safe XML libraries to avoid XML vulnerabilities

XML vulnerabilities are known and well studied. The defusedxml library provides a great synposis of XML vulnerabilities, how they're exploited, and which Python libraries are vulnerable to which attacks.

Most XML vulnerabilities essentially amount to Denial of Service attacks but as previous blackhat presentations have shown, XML vulnerabilities can lead to local file reading, intranet access, and some times remote code execution.

We don't attempt to rehash the details of each vulnerability class and instead recommend those interested read defuxedxml's page, including references.

Incorrect

Currently, the following Python XML libraries are vulnerable to some form of XML attack:

  • xml.sax
    • vulnerable to: billion laughs, quadratic blowup, external entity expansion, DTD retrieval
  • xml.etree.ElementTree
    • vulnerable to: billion laughs, quadratic blowup
  • xml.dom.minidom
    • vulnerable to: billion laughs, quadratic blowup
  • xml.dom.pulldom
    • vulnerable to: billion laughs, quadratic blowup, external entity expansion, DTD retrieval
  • xmlrpclib
    • vulnerable to: billion laughs, quadratic blowup, decompression bomb

Python's XML library page indicates that defusedxml is the correct choice for XML libraries.

Correct

xml.sax

Replace all xml.sax parsers with defusedxml parsers:

  • xml.sax.parser() -> defusedxml.sax.parser()
  • xml.sax.parseString() -> defusedxml.sax.parseString()
  • xml.sax.create_parser() -> defusedxml.sax.parseString()

Intead of this:

    import xml.sax

    class ExampleContentHandler(xml.sax.ContentHandler):
        def __init__(self):
            xml.sax.ContentHandler.__init__(self)

        def startElement(self, name, attrs):
            print 'start:', name

        def endElement(self, name):
            print 'end:', name

        def characters(self, content):
            print 'chars:', content

    def main():
       xml.sax.parse(open('input.xml'), ExampleContentHandler())

    if __name__ == "__main__":
        main()

Do this:

    import xml.sax
    import defusedxml.sax

    class ExampleContentHandler(xml.sax.ContentHandler):
        def __init__(self):
            xml.sax.ContentHandler.__init__(self)

        def startElement(self, name, attrs):
            print 'start:', name

        def endElement(self, name):
            print 'end:', name

        def characters(self, content):
            print 'chars:', content

    def main():
       defusedxml.sax.parse(open('input.xml'), ExampleContentHandler())

    if __name__ == "__main__":
        main()

xml.etree.ElementTree

Replace the following instances of xml.etree.ElementTree functions with the corresponding defusedxml functions:

  • xml.etree.ElementTree.parse() -> defusedxml.ElementTree.parse()
  • xml.etree.ElementTree.iterparse() -> defusedxml.ElementTree.iterparse()
  • xml.etree.ElementTree.fromstring() -> defusedxml.ElementTree.fromstring()
  • xml.etree.ElementTree.XMLParser -> defusedxml.ElementTree.XMLParser

Intead of this:

    import xml.etree.ElementTree as ET
    tree = ET.parse("input.xml")
    root = tree.getroot()

Do this:

    import defusedxml.ElementTree as ET
    tree = ET.parse("input.xml")
    root = tree.getroot()

xml.etree.cElementTree

Replace the following instances of xml.etree.cElementTree functions with the corresponding defusedxml functions:

  • xml.etree.cElementTree.parse() -> defusedxml.cElementTree.parse()
  • xml.etree.cElementTree.iterparse() -> defusedxml.cElementTree.iterparse()
  • xml.etree.cElementTree.fromstring() -> defusedxml.cElementTree.fromstring()
  • xml.etree.cElementTree.XMLParser -> defusedxml.cElementTree.XMLParser

Intead of this:

    import xml.etree.cElementTree as ET
    tree = ET.parse("input.xml")
    root = tree.getroot()

Do this:

    import defusedxml.cElementTree as ET
    tree = ET.parse("input.xml")
    root = tree.getroot()

xml.dom.minidom

Replace the following instances of xml.dom.minidom functions with the corresponding defusedxml functions:

  • xml.dom.minidom.parse() -> defusedxml.minidom.parse()
  • xml.dom.minidom.parseString() -> defusedxml.minidom.parseString()

Intead of this:

    from xml.dom.minidom import parseString
    parseString('<myxml>Some data<empty/> some more data</myxml>')

Do this:

    from defusedxml.minidom import parseString
    parseString('<myxml>Some data<empty/> some more data</myxml>')

xml.dom.pulldom

Replace the following instances of xml.dom.pulldom functions with the corresponding defusedxml functions:

  • xml.dom.pulldom.parse() -> defusedxml.pulldom.parse()
  • xml.dom.pulldom.parseString() -> defusedxml.pulldom.parseString()

Intead of this:

    from xml.dom.pulldom import parseString
    parseString('<myxml>Some data<empty/> some more data</myxml>')

Do this:

    from defusedxml.pulldom import parseString
    parseString('<myxml>Some data<empty/> some more data</myxml>')

xmlrpclib

Taken directly from the defusedxml page: "The function monkey_patch() enables the fixes, unmonkey_patch() removes the patch and puts the code in its former state."

Intead of this:

    from xmlrpclib import ServerProxy, Error

    server = ServerProxy("http://betty.userland.com")

    print server

    try:
        print server.examples.getStateName(41)
    except Error as v:
        print "ERROR", v

Do this:

    from xmlrpclib import ServerProxy, Error
    import defusedxml.xmlrpc

    defusedxml.xmlrpc.monkey_patch()

    server = ServerProxy("http://betty.userland.com")

    print server

    try:
        print server.examples.getStateName(41)
    except Error as v:
        print "ERROR", v

lxml.etree

Replace the following instances of lxml functions with the corresponding defusedxml functions:

  • lxml.etree.parse() -> defusedxml.lxml.parse
  • lxml.etree.fromstring() -> defusedxml.lxml.fromstring()
  • lxml.etree.RestrictedElement() -> defusedxml.lxml.RestrictedElement()
  • lxml.etree.getDefaultParser() -> defusedxml.lxml.getDefaultParser()
  • lxml.etree.check_docinfo() -> defusedxml.lxml.check_docinfo()

Intead of this:

    from lxml import etree
    root = etree.parse('input.xml')

Do this:

    from defusedxml.lxml import parse
    root = parse('input.xml')

References