mirror of https://github.com/CIRCL/AIL-framework
Added new module for Duplicate paste. Seems working but has some small bug (re-check same paste twice)
Mokaddem
parent
0332f23579
commit
14e9850dd6
|
|
@ -0,0 +1,180 @@
|
|||
#!/usr/bin/env python2
|
||||
# -*-coding:UTF-8 -*
|
||||
|
||||
"""
|
||||
The Duplicate module
|
||||
====================
|
||||
|
||||
This huge module is, in short term, checking duplicates.
|
||||
|
||||
Requirements:
|
||||
-------------
|
||||
|
||||
|
||||
"""
|
||||
import redis
|
||||
import os
|
||||
import time
|
||||
import datetime
|
||||
import json
|
||||
import ssdeep
|
||||
from packages import Paste
|
||||
from pubsublogger import publisher
|
||||
from pybloomfilter import BloomFilter
|
||||
|
||||
from Helper import Process
|
||||
|
||||
if __name__ == "__main__":
|
||||
publisher.port = 6380
|
||||
publisher.channel = "Script"
|
||||
|
||||
config_section = 'Duplicates'
|
||||
saved_dico_and_reload = 1 #min
|
||||
time_1 = time.time()
|
||||
flag_reload = True
|
||||
flag_to_disk = False
|
||||
|
||||
p = Process(config_section)
|
||||
|
||||
# REDIS #
|
||||
# DB OBJECT & HASHS ( DISK )
|
||||
# FIXME increase flexibility
|
||||
dico_redis = {}
|
||||
for year in xrange(2013, datetime.date.today().year+1):
|
||||
for month in xrange(0, 16):
|
||||
dico_redis[str(year)+str(month).zfill(2)] = redis.StrictRedis(
|
||||
host=p.config.get("Redis_Level_DB", "host"), port=year,
|
||||
db=month)
|
||||
#print("dup: "+str(year)+str(month).zfill(2)+"\n")
|
||||
|
||||
# FUNCTIONS #
|
||||
publisher.info("Script duplicate started")
|
||||
|
||||
dicopath = os.path.join(os.environ['AIL_HOME'],
|
||||
p.config.get("Directories", "dicofilters"))
|
||||
|
||||
dico_path_set = set()
|
||||
while True:
|
||||
try:
|
||||
hash_dico = {}
|
||||
dupl = []
|
||||
|
||||
x = time.time()
|
||||
|
||||
message = p.get_from_set()
|
||||
if message is not None:
|
||||
path = message
|
||||
PST = Paste.Paste(path)
|
||||
else:
|
||||
publisher.debug("Script Attribute is idling 10s")
|
||||
time.sleep(10)
|
||||
continue
|
||||
|
||||
PST._set_p_hash_kind("ssdeep")
|
||||
|
||||
# Assignate the correct redis connexion
|
||||
r_serv1 = dico_redis[PST.p_date.year + PST.p_date.month]
|
||||
|
||||
# Creating the dicor name: dicoyyyymm
|
||||
filedicopath = os.path.join(dicopath, 'dico' + PST.p_date.year +
|
||||
PST.p_date.month)
|
||||
filedicopath_today = filedicopath
|
||||
|
||||
# Save I/O
|
||||
if time.time() - time_1 > saved_dico_and_reload*60:
|
||||
flag_to_disk = True
|
||||
|
||||
if os.path.exists(filedicopath):
|
||||
if flag_reload == True:
|
||||
flag_reload = False
|
||||
print 'Reloading'
|
||||
time_1 = time.time()
|
||||
with open(filedicopath, 'r') as fp:
|
||||
today_dico = json.load(fp)
|
||||
else:
|
||||
time_1 = time.time()
|
||||
today_dico = {}
|
||||
with open(filedicopath, 'w') as fp:
|
||||
json.dump(today_dico, fp)
|
||||
|
||||
# For now, just use monthly dico
|
||||
dico_path_set.add(filedicopath)
|
||||
|
||||
# UNIQUE INDEX HASHS TABLE
|
||||
yearly_index = str(datetime.date.today().year)+'00'
|
||||
r_serv0 = dico_redis[yearly_index]
|
||||
r_serv0.incr("current_index")
|
||||
index = r_serv0.get("current_index")+str(PST.p_date)
|
||||
# HASHTABLES PER MONTH (because of r_serv1 changing db)
|
||||
r_serv1.set(index, PST.p_path)
|
||||
r_serv1.sadd("INDEX", index)
|
||||
# For each dico
|
||||
opened_dico = []
|
||||
for dico in dico_path_set:
|
||||
# Opening dico
|
||||
if dico == filedicopath_today:
|
||||
opened_dico.append([dico, today_dico])
|
||||
with open(dico, 'r') as fp:
|
||||
opened_dico.append([dico, json.load(fp)])
|
||||
|
||||
|
||||
#retrieve hash from paste
|
||||
paste_hash = PST._get_p_hash()
|
||||
# Adding the hash in Redis
|
||||
r_serv1.set(paste_hash, index)
|
||||
r_serv1.sadd("HASHS", paste_hash)
|
||||
# Go throught the Database of the dico (of the month)
|
||||
threshold_dup = 10
|
||||
for dico_name, dico in opened_dico:
|
||||
for dico_key, dico_hash in dico.items():
|
||||
percent = ssdeep.compare(dico_hash, paste_hash)
|
||||
if percent > threshold_dup:
|
||||
db = dico_name[-6:]
|
||||
# Go throught the Database of the bloom filter (month)
|
||||
r_serv_dico = dico_redis[db]
|
||||
|
||||
# index of paste
|
||||
# FIXME Use r_serv_dico and do not consider only 1 server!!
|
||||
index_current = r_serv1.get(dico_hash)
|
||||
paste_path = r_serv1.get(index_current)
|
||||
if paste_path != None:
|
||||
hash_dico[dico_hash] = (paste_path, percent)
|
||||
|
||||
print 'comparing: ' + str(dico_hash[:20]) + ' and ' + str(paste_hash[:20]) + ' percentage: ' + str(percent)
|
||||
print ' '+ PST.p_path[44:] +', '+ paste_path[44:]
|
||||
|
||||
##################### Similarity found #######################
|
||||
|
||||
# if there is data in this dictionnary
|
||||
if len(hash_dico) != 0:
|
||||
for dico_hash, paste_tuple in hash_dico.items():
|
||||
paste_path, percent = paste_tuple
|
||||
dupl.append((paste_path, percent))
|
||||
|
||||
# Creating the object attribute and save it.
|
||||
to_print = 'Duplicate;{};{};{};'.format(
|
||||
PST.p_source, PST.p_date, PST.p_name)
|
||||
if dupl != []:
|
||||
PST.__setattr__("p_duplicate", dupl)
|
||||
PST.save_attribute_redis("p_duplicate", dupl)
|
||||
publisher.info('{}Detected {}'.format(to_print, len(dupl)))
|
||||
print '{}Detected {}'.format(to_print, len(dupl))
|
||||
|
||||
y = time.time()
|
||||
|
||||
publisher.debug('{}Processed in {} sec'.format(to_print, y-x))
|
||||
|
||||
|
||||
# Adding the hash in the dico of the month
|
||||
today_dico[index] = paste_hash
|
||||
|
||||
if flag_to_disk:
|
||||
flag_to_disk = False
|
||||
flag_reload = True
|
||||
with open(filedicopath, 'w') as fp:
|
||||
json.dump(today_dico, fp)
|
||||
except IOError:
|
||||
to_print = 'Duplicate;{};{};{};'.format(
|
||||
PST.p_source, PST.p_date, PST.p_name)
|
||||
print "CRC Checksum Failed on :", PST.p_path
|
||||
publisher.error('{}CRC Checksum Failed'.format(to_print))
|
||||
|
|
@ -105,7 +105,7 @@ function launching_scripts {
|
|||
|
||||
screen -S "Script" -X screen -t "Global" bash -c './Global.py; read x'
|
||||
sleep 0.1
|
||||
screen -S "Script" -X screen -t "Duplicate" bash -c './Duplicate.py; read x'
|
||||
screen -S "Script" -X screen -t "Duplicate" bash -c './Duplicate_ssdeep.py; read x'
|
||||
sleep 0.1
|
||||
screen -S "Script" -X screen -t "Attribute" bash -c './Attribute.py; read x'
|
||||
sleep 0.1
|
||||
|
|
|
|||
|
|
@ -1,7 +1,7 @@
|
|||
import hashlib
|
||||
import crcmod
|
||||
import mmh3
|
||||
import simhash
|
||||
import ssdeep
|
||||
|
||||
|
||||
class Hash(object):
|
||||
|
|
@ -33,7 +33,7 @@ class Hash(object):
|
|||
elif self.name == "murmur":
|
||||
hash = mmh3.hash(string)
|
||||
|
||||
elif self.name == "simhash":
|
||||
hash = Simhash(string)
|
||||
elif self.name == "ssdeep":
|
||||
hash = ssdeep.hash(string)
|
||||
|
||||
return hash
|
||||
|
|
|
|||
Loading…
Reference in New Issue