misp-taxonomies/tools/machinetag.py

207 lines
11 KiB
Python
Raw Normal View History

#!/usr/bin/env python
# -*- coding: utf-8 -*-
#
# Python script parsing the MISP taxonomies expressed in Machine Tags (Triple
# Tags) to list all valid tags from a specific taxonomy.
#
# Copyright (c) 2015-2022 Alexandre Dulaunoy - a@foo.be
#
# Redistribution and use in source and binary forms, with or without modification,
# are permitted provided that the following conditions are met:
#
# 1. Redistributions of source code must retain the above copyright notice,
# this list of conditions and the following disclaimer.
# 2. Redistributions in binary form must reproduce the above copyright notice,
# this list of conditions and the following disclaimer in the documentation
# and/or other materials provided with the distribution.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
# IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
# INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
# BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
# OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
# OF THE POSSIBILITY OF SUCH DAMAGE.
import json
import os.path
import argparse
import os
import sys
2022-03-25 09:34:14 +01:00
skip_list = ['death-possibilities', 'poison-taxonomy']
taxonomies = []
# Get our current directory from file location
thisDir = os.path.dirname(__file__)
argParser = argparse.ArgumentParser(description='Dump Machine Tags (Triple Tags) from MISP taxonomies', epilog='Available taxonomies are {0}'.format(taxonomies))
argParser.add_argument('-e', action='store_true', help='Include expanded tags')
argParser.add_argument('-a', action='store_true', help='Generate asciidoctor document from MISP taxonomies')
argParser.add_argument('-v', action='store_true', help='Include descriptions')
argParser.add_argument('-n', default=False, help='Show only the specified namespace')
argParser.add_argument('--disable-skip-list', default=False, action='store_true', help='disable default skip list')
args = argParser.parse_args()
if args.disable_skip_list:
skip_list = ''
for folder in os.listdir(os.path.join(thisDir, '../')):
if os.path.isfile(os.path.join(thisDir, '../', folder, 'machinetag.json')):
if folder in skip_list:
continue
taxonomies.append(folder)
taxonomies.sort()
doc = ''
if args.a:
dedication = "\n[dedication]\n== Funding and Support\nThe MISP project is financially and resource supported by https://www.circl.lu/[CIRCL Computer Incident Response Center Luxembourg ].\n\nimage:{images-misp}logo.png[CIRCL logo]\n\nA CEF (Connecting Europe Facility) funding under CEF-TC-2016-3 - Cyber Security has been granted from 1st September 2017 until 31th August 2019 as ***Improving MISP as building blocks for next-generation information sharing***.\n\nimage:{images-misp}en_cef.png[CEF funding]\n\nIf you are interested to co-fund projects around MISP, feel free to get in touch with us.\n\n"
doc = doc + ":toc: right\n"
doc = doc + ":toclevels: 1\n"
doc = doc + ":icons: font\n"
doc = doc + ":images-cdn: https://raw.githubusercontent.com/MISP/MISP/2.4/INSTALL/logos/\n"
doc = doc + ":images-misp: https://www.misp-project.org/assets/images/\n"
doc = doc + "= MISP taxonomies and classification as machine tags\n\n"
doc = doc + "= Introduction\n"
2015-11-29 15:46:13 +01:00
doc = doc + "\nimage::{images-cdn}misp-logo.png[MISP logo]\n"
doc = doc + "The MISP threat sharing platform is a free and open source software helping information sharing of threat intelligence including cyber security indicators, financial fraud or counter-terrorism information. The MISP project includes multiple sub-projects to support the operational requirements of analysts and improve the overall quality of information shared.\n\n"
doc = doc + ""
doc = "{} {} {} {}".format(doc, "\nTaxonomies that can be used in MISP (2.4) and other information sharing tool and expressed in Machine Tags (Triple Tags).",
2017-04-02 22:06:32 +02:00
"A machine tag is composed of a namespace (MUST), a predicate (MUST) and an (OPTIONAL) value.",
"Machine tags are often called triple tag due to their format.\n")
doc = doc + "The following document is generated from the machine-readable JSON describing the https://github.com/MISP/misp-taxonomies[MISP taxonomies]."
doc = doc + "\n\n"
doc = doc + "<<<\n"
doc = doc + dedication
doc = doc + "<<<\n"
doc = doc + "= MISP taxonomies\n"
doc = doc + "\n\n"
if args.n:
del taxonomies[:]
taxonomies.append(args.n)
2017-04-02 22:06:32 +02:00
def asciidoc(content=False, adoc=doc, t='title', toplevel=False):
if not args.a:
return False
adoc = adoc + "\n"
if t == 'title':
content = '==== ' + content
elif t == 'predicate':
content = '=== ' + content
elif t == 'namespace':
content = '== ' + content + '\n'
2020-09-08 11:42:38 +02:00
content = "{}\n{}{} {}{}{} {}".format(content, 'NOTE: ', namespace, 'namespace available in JSON format at https://github.com/MISP/misp-taxonomies/blob/main/',
2017-04-02 22:06:32 +02:00
namespace, '/machinetag.json[*this location*]. The JSON format can be freely reused in your application',
'or automatically enabled in https://www.github.com/MISP/MISP[MISP] taxonomy.')
elif t == 'description' and toplevel is True:
content = "\n{} \n".format(content)
elif t == 'description' and toplevel is False:
try:
(n, value) = content.split(":", 1)
content = "\n{} \n".format(value)
except:
content = "\n{} \n".format(content)
elif t == 'numerical_value':
(n, value) = content.split(":", 1)
content = "\nAssociated numerical value=\"{}\" \n".format(value)
elif t == 'exclusive':
(n, value) = content.split(":", 1)
if n:
content = "\nIMPORTANT: Exclusive flag set which means the values or predicate below must be set exclusively.\n"
adoc = adoc + content
return adoc
2017-04-02 22:06:32 +02:00
def machineTag(namespace=False, predicate=False, value=None):
if namespace is False or predicate is False:
return None
if value is None:
2016-02-11 14:10:58 +01:00
return (u'{0}:{1}'.format(namespace, predicate))
else:
2016-02-11 14:10:58 +01:00
return (u'{0}:{1}=\"{2}\"'.format(namespace, predicate, value))
2017-04-02 22:06:32 +02:00
for taxonomy in taxonomies:
if taxonomy in skip_list:
sys.stderr.write(f"Skip {taxonomy}")
continue
filename = os.path.join(thisDir, "../", taxonomy, "machinetag.json")
with open(filename) as fp:
t = json.load(fp)
namespace = t['namespace']
2016-01-10 17:03:44 +01:00
if t.get('expanded'):
expanded_namespace = t['expanded']
else:
expanded_namespace = namespace
if args.a:
2017-04-02 22:06:32 +02:00
doc = asciidoc(content=t['namespace'], adoc=doc, t='namespace')
doc = asciidoc(content=t['description'], adoc=doc, t='description', toplevel = True)
if t.get('exclusive'):
doc = asciidoc(content=machineTag(namespace=namespace, predicate=t['exclusive']), adoc=doc, t='exclusive')
if args.v:
2017-04-02 22:06:32 +02:00
print('{0}'.format(t['description']))
for predicate in t['predicates']:
if args.a:
doc = asciidoc(content=predicate['value'], adoc=doc, t='predicate')
if predicate.get('description'):
doc = asciidoc(content=machineTag(namespace=namespace, predicate=predicate['description']), adoc=doc, t='description')
if predicate.get('exclusive'):
doc = asciidoc(content=machineTag(namespace=namespace, predicate=predicate['exclusive']), adoc=doc, t='exclusive')
2015-12-01 17:11:43 +01:00
if t.get('values') is None:
if args.a:
doc = asciidoc(content=machineTag(namespace=namespace, predicate=predicate['value']), adoc=doc)
doc = asciidoc(content=machineTag(namespace=namespace, predicate=predicate['expanded']), adoc=doc, t='description')
if predicate.get('description'):
doc = asciidoc(content=machineTag(namespace=namespace, predicate=predicate['description']), adoc=doc, t='description')
if predicate.get('numerical_value'):
doc = asciidoc(content=machineTag(namespace=namespace, predicate=predicate['numerical_value']), adoc=doc, t='description')
if predicate.get('exclusive'):
doc = asciidoc(content=machineTag(namespace=namespace, predicate=predicate['exclusive']), adoc=adoc, t='exclusive')
else:
2017-04-02 22:06:32 +02:00
print(machineTag(namespace=namespace, predicate=predicate['value']))
if args.e:
2017-04-02 22:06:32 +02:00
print("--> " + machineTag(namespace=expanded_namespace, predicate=predicate['expanded']))
if predicate.get('description'):
print("--> " + predicate['description'])
else:
for e in t['values']:
if e['predicate'] == predicate['value']:
if 'expanded' in predicate:
expanded = predicate['expanded']
for v in e['entry']:
2016-06-21 07:53:31 +02:00
if args.a and 'expanded' in v:
doc = asciidoc(content=machineTag(namespace=namespace, predicate=e['predicate'], value=v['value']), adoc=doc)
doc = asciidoc(content=machineTag(namespace=namespace, predicate=v['expanded']), adoc=doc, t='description')
if 'description' in v:
doc = asciidoc(content=machineTag(namespace=namespace, predicate=v['description']), adoc=doc, t='description')
if v.get('numerical_value'):
doc = asciidoc(content=machineTag(namespace=namespace, predicate=v['numerical_value']), adoc=doc, t='numerical_value')
else:
2017-04-02 22:06:32 +02:00
print(machineTag(namespace=namespace, predicate=e['predicate'], value=v['value']))
if args.e:
2016-06-21 07:35:55 +02:00
if'expanded' in v:
2017-04-02 22:06:32 +02:00
print("--> " + machineTag(namespace=namespace, predicate=expanded, value=v['expanded']))
with open('../mapping/mapping.json') as mapping:
m = json.load(mapping)
output = '\n= Mapping of taxonomies\n'
output = '{}{}'.format(output, 'Analysts relying on taxonomies don\'t always know the appropriate namespace to use but know which value to use for classification. The MISP mapping taxonomy allows to map a single classification into a series of machine-tag synonyms.\n')
for value in sorted(m.keys()):
output = '{}{} **{}**{}{}\n'.format(output,'\n.Mapping table - ',value,'\n|===\n|',value)
for mapped in m[value]['values']:
output = '{}|{}\n'.format(output,mapped)
output = '{}|===\n'.format(output)
doc = doc + output
if args.a:
2017-04-02 22:06:32 +02:00
print(doc)