ail-framework/bin/modules/CEDetector.py
terrtia e55aeab11c
Some checks failed
CI / ail_test (3.10) (push) Has been cancelled
CI / ail_test (3.7) (push) Has been cancelled
CI / ail_test (3.8) (push) Has been cancelled
CI / ail_test (3.9) (push) Has been cancelled
chg: [module] debug
2024-10-11 14:55:04 +02:00

130 lines
3.8 KiB
Python
Executable file

#!/usr/bin/env python3
# -*-coding:UTF-8 -*
"""
The Onion Module
============================
This module extract url from item and returning only ones which are tor
related (.onion). All These urls are send to the crawler discovery queue.
Requirements
------------
*Need running Redis instances. (Redis)
"""
import os
import sys
from textblob import TextBlob
from nltk.tokenize import RegexpTokenizer
sys.path.append(os.environ['AIL_BIN'])
##################################
# Import Project packages
##################################
from modules.abstract_module import AbstractModule
from lib.ConfigLoader import ConfigLoader
class CEDetector(AbstractModule):
"""docstring for Onion module."""
def __init__(self, queue=True):
super(CEDetector, self).__init__(queue=queue)
config_loader = ConfigLoader()
self.r_cache = config_loader.get_redis_conn("Redis_Cache")
self.csam_words = self.load_world_file('csam_words')
self.child_worlds = self.load_world_file('child_words')
self.porn_worlds = self.load_world_file('porn_words')
self.ce_tag = 'dark-web:topic="pornography-child-exploitation"'
self.tokenizer = RegexpTokenizer('[\&\~\:\;\,\.\(\)\{\}\|\[\]\\\\//\=\'\"\%\$\?\@\+\#\_\^\<\>\!\*\n\r\t\s]+',
gaps=True, discard_empty=True)
def load_world_file(self, path):
words = set()
try:
with open(os.path.join(os.environ['AIL_HOME'], f'files/{path}')) as f:
content = f.read()
except FileNotFoundError:
content = ''
content = content.splitlines()
for line in content:
if line.startswith('#') or not line:
continue
word = line.split()
if word:
words.add(word[0])
return words
def compute(self, message): # TODO LIMIT TO DARKWEB ???
to_tag = False
content = self.obj.get_content().lower()
# print(content)
is_csam = False
is_child_word = False
is_porn_world = False
words = TextBlob(content, tokenizer=self.tokenizer).tokens
words = set(words)
for word in words:
if word in self.csam_words:
is_csam = True
if word in self.child_worlds:
is_child_word = True
if word in self.porn_worlds:
is_porn_world = True
# PERF ???
# if is_child_word and is_porn_world:
# break
if is_csam:
to_tag = True
if is_child_word and is_porn_world:
to_tag = True
if to_tag:
# print(f'{content} DETECTED')
# print()
self.add_message_to_queue(message=self.ce_tag, queue='Tags')
return to_tag
def test_detection():
from lib import Tag
from lib.objects.Domains import Domain
from lib.objects.Titles import Title
not_detected = set()
tag = 'dark-web:topic="pornography-child-exploitation"'
tag_key = f'domain::{tag}'
for domain in Tag.get_obj_by_tag(tag_key):
dom = Domain(domain)
is_detected = False
for h in dom.get_correlation('title').get('title', []):
module.obj = Title(h[1:])
if module.compute(''):
is_detected = True
if not is_detected:
not_detected.add(domain)
print()
print()
print()
print()
for domain in not_detected:
dom = Domain(domain)
print('-----------', domain)
for h in dom.get_correlation('title').get('title', []):
print(Title(h[1:]).get_content().lower())
print()
print()
if __name__ == "__main__":
module = CEDetector()
module.run()
# test_detection()