mirror of
https://github.com/MISP/misp-galaxy.git
synced 2024-11-26 00:37:18 +00:00
wip: [GSMA MoTIF] new galaxy - initial pdf parsing
This commit is contained in:
parent
20ff10b5b1
commit
63b7725cfb
3 changed files with 227 additions and 0 deletions
9
.vscode/launch.json
vendored
9
.vscode/launch.json
vendored
|
@ -1,6 +1,15 @@
|
|||
{
|
||||
"version": "0.2.0",
|
||||
"configurations": [
|
||||
{
|
||||
"name": "gen_gsma_motif",
|
||||
"type": "debugpy",
|
||||
"request": "launch",
|
||||
"program": "${file}",
|
||||
"console": "integratedTerminal",
|
||||
"args": "",
|
||||
"cwd": "${fileDirname}"
|
||||
},
|
||||
{
|
||||
"name": "gen_mitre_d3fend",
|
||||
"type": "debugpy",
|
||||
|
|
213
tools/gen_gsma_motif.py
Normal file
213
tools/gen_gsma_motif.py
Normal file
|
@ -0,0 +1,213 @@
|
|||
#!/usr/bin/env python3
|
||||
#
|
||||
# A simple convertor of the GSMA Mobile Threat Intelligence Framework (MoTIF) Principles to a MISP Galaxy datastructure.
|
||||
# https://www.gsma.com/security/resources/fs-57-mobile-threat-intelligence-framework-motif-principles/
|
||||
# Copyright (c) 2024 MISP Project
|
||||
# Copyright (c) 2024 Christophe Vandeplas
|
||||
#
|
||||
# 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 pdfplumber
|
||||
import requests
|
||||
import json
|
||||
|
||||
|
||||
pdf_file_name = 'gsma-motif-principles.pdf'
|
||||
pdf_url = 'https://www.gsma.com/solutions-and-impact/technologies/security/wp-content/uploads/2024/04/FS.57-MoTIF-Principles-v1.0.pdf'
|
||||
|
||||
# FIXME save to temporary file
|
||||
# r = requests.get(pdf_url, allow_redirects=True)
|
||||
# pdf_file_content = r.content
|
||||
|
||||
# with open(pdf_file_name, 'wb') as f:
|
||||
# f.write(pdf_file_content)
|
||||
|
||||
'''
|
||||
- search for 4.1 Techniques and Sub-techniques Definition
|
||||
- this is the start of our table parsing stuff
|
||||
- each title 4.1.x is a new technique - MOTxxxx
|
||||
- each title 4.1.x.y is a new sub-technique - MOTxxxx.yyy
|
||||
- extract table,
|
||||
-- identify if table is finishing on next page
|
||||
-- if so jump to next page and merge both tables
|
||||
- make generic function to convert table to dict
|
||||
|
||||
'''
|
||||
|
||||
|
||||
def sub_table_to_list(table: list) -> list:
|
||||
if len(table) == 0:
|
||||
return []
|
||||
try:
|
||||
result = []
|
||||
# FIXME use header row to know column names
|
||||
for row in table:
|
||||
result.append({
|
||||
'ID': row[2].replace('\n', ''),
|
||||
'Name': row[4]. replace('\n', ' ').strip(),
|
||||
'Description': row[5]
|
||||
})
|
||||
return result
|
||||
except IndexError:
|
||||
return []
|
||||
|
||||
|
||||
def table_to_technique(table: list) -> dict:
|
||||
'''
|
||||
Convert a table to a technique dictionary
|
||||
'''
|
||||
result = {}
|
||||
row_index = 0
|
||||
while row_index < len(table):
|
||||
row = table[row_index]
|
||||
|
||||
# row[1] is None : sub-table in table
|
||||
field = cleanup_field(row[0])
|
||||
try:
|
||||
if result['ID'] == 'MOT1036.301':
|
||||
pass
|
||||
except KeyError:
|
||||
pass
|
||||
if field == 'Procedure Examples':
|
||||
# extract sub-table in the next rows
|
||||
sub_table = []
|
||||
try:
|
||||
while table[row_index + 1][0] is None:
|
||||
sub_table.append(table[row_index + 1])
|
||||
row_index += 1
|
||||
except IndexError: # just the end of the page, will be handled in the next page
|
||||
pass
|
||||
value = sub_table_to_list(sub_table)
|
||||
elif field == 'Analogous technique in other frameworks':
|
||||
# column index is not always the same... so figure out the first non-empty cell
|
||||
i = 1
|
||||
value = ''
|
||||
while i < len(row):
|
||||
try:
|
||||
if row[i] is not None:
|
||||
value = row[i]
|
||||
break
|
||||
except IndexError:
|
||||
pass
|
||||
i += 1
|
||||
elif not field:
|
||||
# annoyingly a sub-table might have been parsed differently from previous page. So bad luck. There's not much we can do about it except even worse code than we have here.
|
||||
row_index += 1
|
||||
continue
|
||||
else:
|
||||
value = row[1].replace('\n', ' ').strip()
|
||||
|
||||
result[field] = value
|
||||
|
||||
row_index += 1
|
||||
|
||||
return result
|
||||
|
||||
|
||||
def cleanup_field(field: str) -> str:
|
||||
'''
|
||||
Cleanup a field name
|
||||
'''
|
||||
try:
|
||||
return field.strip().replace(':', '').replace('\n', ' ').replace('- ', '-').strip()
|
||||
except AttributeError:
|
||||
return ''
|
||||
|
||||
|
||||
def is_end_of_table(table: list) -> bool:
|
||||
'''
|
||||
Check if this is the end of the table, by checking the last row in the table.
|
||||
'''
|
||||
try:
|
||||
# Techniques
|
||||
if table['ID'].startswith('MOT') and 'Analogous technique in other frameworks' in table:
|
||||
return True
|
||||
# Mitigations
|
||||
if table['ID'].startswith('MOS') and 'References' in table:
|
||||
return True
|
||||
|
||||
except KeyError:
|
||||
pass
|
||||
return False
|
||||
|
||||
|
||||
entries = {}
|
||||
|
||||
table_settings = {
|
||||
"vertical_strategy": "lines",
|
||||
"horizontal_strategy": "lines",
|
||||
# "explicit_vertical_lines": [],
|
||||
# "explicit_horizontal_lines": [],
|
||||
# "snap_tolerance": 6,
|
||||
"snap_x_tolerance": 6, # pg49: must be 6
|
||||
"snap_y_tolerance": 3, # max 14
|
||||
# "join_tolerance": 3,
|
||||
# "join_x_tolerance": 3,
|
||||
# "join_y_tolerance": 3,
|
||||
# "edge_min_length": 3,
|
||||
# "min_words_vertical": 3,
|
||||
# "min_words_horizontal": 1,
|
||||
# "intersection_tolerance": 3,
|
||||
# "intersection_x_tolerance": 3,
|
||||
# "intersection_y_tolerance": 3,
|
||||
# "text_tolerance": 3,
|
||||
# "text_x_tolerance": 3,
|
||||
# "text_y_tolerance": 3,
|
||||
}
|
||||
|
||||
with pdfplumber.open(pdf_file_name) as pdfp:
|
||||
|
||||
page_index = 0
|
||||
title_seen = False
|
||||
curr_table = None
|
||||
|
||||
while page_index < len(pdfp.pages):
|
||||
page = pdfp.pages[page_index]
|
||||
# skip to section 4.1 Techniques and Sub-techniques Definition
|
||||
if not title_seen:
|
||||
page_text = page.extract_text()
|
||||
if '4.1 Techniques and Sub-techniques Definition' not in page_text or 'Table of Contents' in page_text:
|
||||
print(f"Skipping page {page_index}")
|
||||
page_index += 1
|
||||
continue
|
||||
title_seen = True
|
||||
|
||||
# parse technique tables
|
||||
|
||||
for table in page.extract_tables(table_settings=table_settings):
|
||||
if curr_table: # merge tables if continuation
|
||||
# if first row does not have a first column, then it's the continuation of the previous row
|
||||
if table[0][0] == '' and table[0][1] != '':
|
||||
curr_table[-1][1] += ' ' + table[0][1] # add description of new row to previous row
|
||||
table.pop(0) # remove the first new row of the table
|
||||
# annoyingly a sub-table might have been parsed differently from previous page. So bad luck. There's not much we can do about it except even worse code than we have here.
|
||||
# handle rest of merging case
|
||||
table = curr_table + table
|
||||
curr_table = None # reset for clean start
|
||||
|
||||
parsed_table = table_to_technique(table)
|
||||
if is_end_of_table(parsed_table):
|
||||
# valid table
|
||||
entries[parsed_table['ID']] = parsed_table
|
||||
if parsed_table['ID'] == 'MOT1565.001':
|
||||
pass
|
||||
else:
|
||||
# incomplete table, store in curr_table and continue next row
|
||||
curr_table = table
|
||||
page_index += 1
|
||||
|
||||
with open('gsma_motif.json', 'w') as f:
|
||||
json.dump(entries, f, indent=4)
|
||||
|
5
tools/requirements.txt
Normal file
5
tools/requirements.txt
Normal file
|
@ -0,0 +1,5 @@
|
|||
|
||||
pdfplumber==0.11.0
|
||||
graphviz==0.20.3
|
||||
requests==2.23.3
|
||||
|
Loading…
Reference in a new issue