misp-galaxy/tools/gen_relationships.py

212 lines
11 KiB
Python
Raw Normal View History

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
#
#
# Facilitates the creation and maintenance of relationships.
# Copyright (C) 2022 MISP Project
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License as
# published by the Free Software Foundation, either version 3 of the
# License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU Affero General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
import argparse
import os
import json
import logging
import re
def relation_exists(cluster_a, cluster_b):
"""
Checks if there is already a relationship from a to b.
Note: you might want to run this function from a to b and from b to a.
"""
try:
for rel in cluster_a['related']:
if cluster_b['uuid'] == rel['dest-uuid']:
return True
except KeyError: # no relations yet
pass
return False
def create_relation(cluster_a, cluster_b, rel_type="similar", tags=None):
"""
Creates unidirectional relationship, with a (optional) tags
"""
if not relation_exists(cluster_a, cluster_b):
rel = {"dest-uuid": cluster_b['uuid'],
"type": rel_type}
if tags:
rel["tags"] = tags
if 'related' not in cluster_a:
cluster_a['related'] = []
cluster_a['related'].append(rel)
return True
return False
class AtLeastTwoItemsAction(argparse.Action):
def __call__(self, parser, namespace, values, option_string=None):
if len(values) < 2:
parser.error('argument "{}" requires at least 2 values'.format(self.dest))
else:
filenames = []
for value in values:
if not os.path.isfile(value):
parser.error('file "{}" does not exist'.format(value))
filenames.append(value)
setattr(namespace, self.dest, filenames)
parser = argparse.ArgumentParser(description="MISP Galaxy relationship tool.")
parser.add_argument("files", nargs='+',
help="The names of the clusters. (filename or cluster-name)",
action=AtLeastTwoItemsAction)
parser.add_argument("-ss", "--synonyms-source",
help="Also use synonyms from the source cluster",
action='store_true')
parser.add_argument("-sd", "--synonyms-destination",
help="Also use synonyms from the destination cluster from which we are looking up",
action='store_true')
parser.add_argument("-y", "--yes",
help="Assume yes to all the questions, so create relationships without asking.",
action='store_true') # FIXME develop this feature
parser.add_argument('-v', '--verbose', action='count', default=0)
args = parser.parse_args()
levels = [logging.WARNING, logging.INFO, logging.DEBUG]
level = levels[min(args.verbose, len(levels) - 1)] # cap to last level index
logging.basicConfig(level=level, format="%(message)s")
cluster_files = {}
cluster_files_changed_tracking = {}
# load all non-deprecated Clusters in memory
logging.info("Reading all non-deprecated cluster files in memory.")
for filename in args.files:
# skip if deprecated
galaxy_filename = os.path.join(os.path.dirname(filename), '..', 'galaxies', os.path.basename(filename))
with open(galaxy_filename, 'r') as f:
f_content = json.load(f)
if f_content.get('namespace') == 'deprecated':
logging.debug(f"Skipping file {filename} as Galaxy is deprecated.")
continue
# load file in memory
logging.debug(f"Loading {filename} in memory.")
with open(filename, 'r') as f:
f_content = json.load(f)
cluster_files[filename] = f_content
cluster_files_changed_tracking[filename] = False
logging.info("Processing clusters.")
# tags by default
if args.synonyms_destination or args.synonyms_source:
tags = ["estimative-language:likelihood-probability=\"likely\""]
else:
tags = ["estimative-language:likelihood-probability=\"almost-certain\""]
rel_type = 'similar'
# process each cluster one by one
try:
for cluster_filename, clusters in cluster_files.items():
logging.debug(f"Processing cluster {cluster_filename}.")
for cluster in clusters['values']:
values_to_lookup = [cluster['value'].lower()]
if args.synonyms_source:
try:
values_to_lookup.extend([value.lower() for value in cluster['meta']['synonyms']])
except KeyError:
pass
for lookup_cluster_filename, lookup_clusters in cluster_files.items():
if lookup_cluster_filename == cluster_filename: # skip current cluster
continue
for lookup_cluster in lookup_clusters['values']:
lookup_cluster_values = [lookup_cluster['value'].lower()]
if args.synonyms_destination:
try:
lookup_cluster_values.extend(
[value.lower() for value in lookup_cluster['meta']['synonyms']])
except KeyError:
pass
if any(item in values_to_lookup for item in lookup_cluster_values):
# we have a match from any of our source strings in the lookup cluster
if not relation_exists(cluster, lookup_cluster): # no relation yet, create it
if args.yes:
logging.info(f"Found non-existing match for {cluster_filename} {values_to_lookup} in {lookup_cluster_filename} {lookup_cluster_values}. Creating it")
create_relationship = True
else:
# interactive prompt to ask what to do
# TODO add question which tags and relationship type and more
print(f"Found non-existing match for {cluster_filename} {values_to_lookup} in {lookup_cluster_filename} {lookup_cluster_values}.")
while True:
user_input = input(f"Create relation? [\u0332yes] / \u0332no / \u0332details / \u0332tags / \u0332relation: ").lower().strip()
if user_input in ['yes', 'y', '']:
create_relationship = True
logging.info(" creating it.")
break
if user_input in ['no', 'n']:
create_relationship = False
break
if user_input in ['tags', 'tag', 't']:
tags = sorted(tags)
while True:
print(f"Current tags: ")
[print(f" [{i}]: {t}") for i, t in enumerate(tags)]
tag_input = input(f"Change tags? [\u0332no] / \u0332add / \u0332delete #: ").lower().strip()
if tag_input in ['n', 'no', '']:
break
if tag_input.startswith('delete ') or tag_input.startswith('d '):
try:
tag_delete_ids = [int(n) for n in tag_input.split(' ')[1:]]
tag_delete_ids.sort(reverse=True)
for i in tag_delete_ids:
del tags[i]
except (ValueError, IndexError):
pass
if tag_input in ['add', 'a']:
new_tag = input("Enter tag to add: ").strip()
if re.match(r'[^:]+:[^:]+="[^"]+"', new_tag) or re.match(r'[\w]+:[\w]+', new_tag):
tags.append(new_tag)
else:
print("ERROR: Tag is not in the proper structure.")
if user_input in ['relation', 'r']:
new_relation = input(f"Current relation is '{rel_type}'. Enter new: ").lower().strip()
if new_relation:
rel_type = new_relation
if user_input in ['details', 'd']:
print("Is:")
print(f" {cluster_filename} with values: {values_to_lookup}:")
print(f" {cluster.get('description')}")
print(f"{rel_type}:")
print(f" {lookup_cluster_filename} with values: {lookup_cluster_values}:")
print(f" {lookup_cluster.get('description')}")
print("")
print(f"Tags: {tags}")
if create_relationship:
if create_relation(cluster, lookup_cluster, rel_type=rel_type, tags=tags):
cluster_files_changed_tracking[cluster_filename] = True
if create_relation(lookup_cluster, cluster, rel_type=rel_type, tags=tags):
cluster_files_changed_tracking[lookup_cluster_filename] = True
except KeyboardInterrupt:
pass
# save all to file, and increment version number if something changed
for cluster_filename, changed in cluster_files_changed_tracking.items():
if not changed:
continue
logging.debug(f"File {cluster_filename} has changed. Saving...")
with open(cluster_filename, 'w') as f:
cluster_files[cluster_filename]['version'] += 1
json.dump(cluster_files[cluster_filename], f, indent=2, sort_keys=True, ensure_ascii=False)
f.write('\n') # only needed for the beauty and to be compliant with jq_all_the_things
print("All done, please don't forget to ./jq_all_the_things.sh, commit, and then ./validate_all.sh.")