def __init__(self):
# load up the tld list now as only hit it once this way
self.pubsuffix_list = self.get_pubsuffix_list()
# this is to speed up tlds lookup with hash table
# we can share among all runs over time
self.sql_driver = MySQLDriver()
# check if sub_domain_tld exists, if not create one
# this should only really happen once
#
# bug?
# when using pool this can be done several times in tandem
# resulting in some lost entries - e.g. two threads see db not exist the first
# one to break the tie will create a new blank db and add a new entry
# it is possible the slightly slower second thread will also create a blank db
# erasing the entry of the first thread, however, after the first round the db
# should not be re-created again
#
# it should be emphasized this db is ONLY used to speed up parsing the tld
# (partially because my regex method is slow and should be refactored!)
# the data is this db is NOT used for analysis - thus deleting a few records
# will have a minor impact on speed for the first few entries, and then be
# wholly irrelevant
#
# still...it irks me that this entire class could be smarter, so I will
# fix it another day
if self.sql_driver.check_db_exist('sub_domain_tld') == False:
self.sql_driver.create_sub_domain_tld_db()
else:
self.sql_driver.db_switch('sub_domain_tld')
def __init__(self, db_name, num_tlds, num_results, tracker_threshold=0):
self.db_name = db_name
self.sql_driver = MySQLDriver(self.db_name)
self.num_tlds = num_tlds
self.num_results = num_results
self.tracker_threshold = tracker_threshold
self.startTime = datetime.now()
self.pages_ok_count = self.sql_driver.pages_ok_count()
print('\t=============================')
print('\t Checking Output Directories ')
print('\t=============================')
self.setup_report_dir()
print('\t===========================')
print('\t Patching DB with Org Data ')
print('\t===========================')
# update the domains to their owners
self.patch_org_data()
print('\t\tSuccess!')
print('\t=====================')
print('\t Getting top %s tlds' % self.num_tlds)
print('\t=====================')
print('\t\tProcessing...')
self.top_tlds = self.get_top_tlds(self.num_tlds)
print(self.top_tlds)
print('\t\tSuccess!')
print('\t\tThe top tlds are:')
for (tld, pages) in self.top_tlds:
print('\t\t |- %s (%s)' % (tld, pages))
# SPECIAL SAUCE, FOR EXPERTS: tracker domains!
#
# idea for this is you set a threshold of the number of sites a given domain
# is connected to - domains connecting to many sites may correlate those visits
# via referer strings etc, so we call these 'tracker domains'
#
# on a really large set of sites (e.g. 1M+) this works well but on small samples
# (e.g. 500) it doesn't work well at all as known tracker domains may only
# appear on a single site
#
# this is off by default and unless you understand what you are doing...
# DON'T USE THIS!
#
# longer-term we may want to train off a bigger corpus to find tracker domains and
# have them prepackaged
#
if tracker_threshold:
print('\t=========================')
print('\t Getting tracker domains ')
print('\t=========================')
print('\t\tProcessing...')
self.tracker_domains = self.get_tracker_domains(
self.tracker_threshold)
print('\t\tSuccess!')
else:
self.tracker_domains = []
def run(self, pool_size):
try:
uri_list = open('./page_lists/' + self.pages_file_name, 'r')
except:
print(
'File "%s" does not exist, file must be in ./page_lists directory. Exiting.'
% self.pages_file_name)
exit()
sql_driver = MySQLDriver(self.db_name)
# sort out what uris we are processing from the list
uris_to_process = []
count = 0
print('\t------------------------')
print('\t Building List of Pages ')
print('\t------------------------')
for uri in uri_list:
# skip lines that are comments
if "#" in uri[0]: continue
count += 1
# drop trailing '/, clean off white space, make lower, create cli-safe uri
# with parse.quote, but exclude :/ b/c of http://
uri = re.sub('/$', '',
urllib.parse.quote(uri.strip(), safe=":/").lower())
# if it is a m$ office or other doc, skip
if re.match('.+(pdf|ppt|pptx|doc|docx|txt|rtf|xls|xlsx)$', uri):
print("\t\t%s | %-50s Not an HTML document, Skipping." %
(count, uri[:50]))
continue
# skip if in db already
if sql_driver.page_exists(uri):
print("\t\t%s | %-50s Exists in DB, Skipping." %
(count, uri[:50]))
continue
# only add if not in list already
if uri not in uris_to_process:
print("\t\t%s | %-50s Adding." % (count, uri[:50]))
uris_to_process.append(uri)
else:
print("\t\t%s | %-50s Already queued, Skipping." %
(count, uri[:50]))
print('\t----------------------------------')
print('\t%s pages will now be webXray\'d' % len(uris_to_process))
print('\t\t...you can go take a walk. ;-)')
print('\t----------------------------------')
myPool = Pool(pool_size)
myPool.map(self.process_uri, uris_to_process)
def run(self, pool_size):
try:
uri_list = open('./page_lists/'+self.pages_file_name, 'r')
except:
print('File "%s" does not exist, file must be in ./page_lists directory. Exiting.' % self.pages_file_name)
exit()
sql_driver = MySQLDriver(self.db_name)
# sort out what uris we are processing from the list
uris_to_process = []
count = 0
print('\t------------------------')
print('\t Building List of Pages ')
print('\t------------------------')
for uri in uri_list:
# skip lines that are comments
if "#" in uri[0]: continue
count += 1
# only do lines starting with https?://
if not (re.match('^https?://.+', uri)):
print("\t\t%s | %-50s Not a valid address, Skipping." % (count, uri[:50]))
continue
# drop trailing '/, clean off white space, make lower, create cli-safe uri
# with parse.quote, but exclude :/ b/c of http://
uri = re.sub('/$', '', urllib.parse.quote(uri.strip(), safe=":/").lower())
# if it is a m$ office or other doc, skip
if re.match('.+(pdf|ppt|pptx|doc|docx|txt|rtf|xls|xlsx)$', uri):
print("\t\t%s | %-50s Not an HTML document, Skipping." % (count, uri[:50]))
continue
# skip if in db already
if sql_driver.page_exists(uri):
print("\t\t%s | %-50s Exists in DB, Skipping." % (count, uri[:50]))
continue
# only add if not in list already
if uri not in uris_to_process:
print("\t\t%s | %-50s Adding." % (count, uri[:50]))
uris_to_process.append(uri)
else:
print("\t\t%s | %-50s Already queued, Skipping." % (count, uri[:50]))
print('\t----------------------------------')
print('\t%s addresses will now be webXray\'d' % len(uris_to_process))
print('\t\t...you can go take a walk. ;-)')
print('\t----------------------------------')
myPool = Pool(pool_size)
myPool.map(self.process_uri, uris_to_process)
def process_uri(self, uri):
sql_driver = MySQLDriver(self.db_name)
output_store = OutputStore(self.db_name)
phantom_driver = PhantomDriver(
'--ignore-ssl-errors=true --ssl-protocol=any', 'wbxr_logger.js')
# this can be higher or lower depending on network load
# generally, 90 seems to be fine, so keep with it
try:
phantom_output = phantom_driver.execute(uri, 90)
except:
print("\t\t%-50s Phantomjs Did Not Return." % uri[:50])
sql_driver.log_error(uri, "FAIL: Phantomjs Did Not Return.")
return
if re.match('^FAIL.+', phantom_output):
print("\t\t%-50s Phantom Error\n\t%s" % (uri[:50], phantom_output))
sql_driver.log_error(uri, phantom_output)
else:
print("\t\t%-50s %s" %
(uri[:50], output_store.store(uri, phantom_output)))
# closes our db connections
sql_driver.close()
output_store.close()
return
def __init__(self):
# load up the tld list now as only hit it once this way
self.pubsuffix_list = self.get_pubsuffix_list()
# this is to speed up tlds lookup with hash table
# we can share among all runs over time
self.sql_driver = MySQLDriver()
# check if sub_domain_tld exists, if not create one
# this should only really happen once
#
# bug?
# when using pool this can be done several times in tandem
# resulting in some lost entries - e.g. two threads see db not exist the first
# one to break the tie will create a new blank db and add a new entry
# it is possible the slightly slower second thread will also create a blank db
# erasing the entry of the first thread, however, after the first round the db
# should not be re-created again
#
# it should be emphasized this db is ONLY used to speed up parsing the tld
# (partially because my regex method is slow and should be refactored!)
# the data is this db is NOT used for analysis - thus deleting a few records
# will have a minor impact on speed for the first few entries, and then be
# wholly irrelevant
#
# still...it irks me that this entire class could be smarter, so I will
# fix it another day
if self.sql_driver.check_db_exist('sub_domain_tld') == False:
self.sql_driver.create_sub_domain_tld_db()
else:
self.sql_driver.db_switch('sub_domain_tld')
def __init__(self, db_name, num_tlds, num_results, tracker_threshold = 0):
self.db_name = db_name
self.sql_driver = MySQLDriver(self.db_name)
self.num_tlds = num_tlds
self.num_results = num_results
self.tracker_threshold = tracker_threshold
self.startTime = datetime.now()
self.pages_ok_count = self.sql_driver.pages_ok_count()
print('\t=============================')
print('\t Checking Output Directories ')
print('\t=============================')
self.setup_report_dir()
print('\t===========================')
print('\t Patching DB with Org Data ')
print('\t===========================')
# update the domains to their owners
self.patch_org_data()
print('\t\tSuccess!')
print('\t=====================')
print('\t Getting top %s tlds' % self.num_tlds)
print('\t=====================')
print('\t\tProcessing...')
self.top_tlds = self.get_top_tlds(self.num_tlds)
print(self.top_tlds)
print('\t\tSuccess!')
print('\t\tThe top tlds are:')
for (tld, pages) in self.top_tlds:
print('\t\t |- %s (%s)' % (tld,pages))
# SPECIAL SAUCE, FOR EXPERTS: tracker domains!
#
# idea for this is you set a threshold of the number of sites a given domain
# is connected to - domains connecting to many sites may correlate those visits
# via referer strings etc, so we call these 'tracker domains'
#
# on a really large set of sites (e.g. 1M+) this works well but on small samples
# (e.g. 500) it doesn't work well at all as known tracker domains may only
# appear on a single site
#
# this is off by default and unless you understand what you are doing...
# DON'T USE THIS!
#
# longer-term we may want to train off a bigger corpus to find tracker domains and
# have them prepackaged
#
if tracker_threshold:
print('\t=========================')
print('\t Getting tracker domains ')
print('\t=========================')
print('\t\tProcessing...')
self.tracker_domains = self.get_tracker_domains(self.tracker_threshold)
print('\t\tSuccess!')
else:
self.tracker_domains = []
async def post_analyze_url(request: XrayAnalyseRequest):
print('Data recieve.')
status = False
cookies = []
driver = MySQLDriver('wbxr_gayatri')
try:
url = unpadPKCS7(decrypt(request.url))
url = url.decode('utf8')
print('Encoded', url)
if history_filter(url):
driver.db.execute(
"SELECT cookie.`domain` "
"FROM page "
"LEFT JOIN page_cookie_junction "
"ON page.id = page_cookie_junction.page_id "
"LEFT JOIN cookie "
"ON page_cookie_junction.cookie_id = cookie.id "
"WHERE page.start_uri_md5 = MD5(%s)",
(url,),
)
fetched = driver.db.fetchall()
if fetched:
cookies = fetched
else:
cookies = analyze_url(url)
else:
cookies = []
print(cookies)
except Exception as e:
print(e)
else:
status = True
return JSONResponse(content={"status": status, "cookies": cookies})
def process_uri(self, uri):
sql_driver = MySQLDriver(self.db_name)
output_store = OutputStore(self.db_name)
phantom_driver = PhantomDriver('--ignore-ssl-errors=true --ssl-protocol=any', 'wbxr_logger.js')
# this can be higher or lower depending on network load
# generally, 90 seems to be fine, so keep with it
try:
phantom_output = phantom_driver.execute(uri, 90)
except:
print("\t\t%-50s Phantomjs Did Not Return." % uri[:50])
sql_driver.log_error(uri, "FAIL: Phantomjs Did Not Return.")
return
if re.match('^FAIL.+', phantom_output):
print("\t\t%-50s Phantom Error\n\t%s" % (uri[:50], phantom_output))
sql_driver.log_error(uri, phantom_output)
else:
print("\t\t%-50s %s" % (uri[:50], output_store.store(uri, phantom_output)))
# closes our db connections
sql_driver.close()
output_store.close()
return
if sys.version_info[1] < 4:
print(
'Python 3.4 or above is required for webXray to function; please check your installation.'
)
exit()
# standard python 3.4 libs
import os
import re
import time
from optparse import OptionParser
# set up a global mysql driver, in the future you could use other db drivers here
# if the mysql connector is not installed this fails gracefully
from webxray.MySQLDriver import MySQLDriver
sql_driver = MySQLDriver()
# databases are stored with a 'wbxr_' prefix, this function helps select a database in interactive mode
def select_wbxr_db():
wbxr_dbs = sql_driver.get_wbxr_dbs_list()
if len(wbxr_dbs) == 0:
print(
'''\t\tThere are no databases to analyze, please try [C]ollecting data or
import an existing wbxr-formatted database manually.''')
interaction()
return
for index, db_name in enumerate(wbxr_dbs):
print('\t\t[%s] %s' % (index, db_name[5:]))
def __init__(self, dbname): self.uri_parser = ParseURI() self.sql_driver = MySQLDriver(dbname)
class OutputStore:
def __init__(self, dbname):
self.uri_parser = ParseURI()
self.sql_driver = MySQLDriver(dbname)
# end init
def store(self, uri, phantom_output):
# parse out the json from our phantom_output
# sometimes phantom prints out errors before the json, (despite us turning
# it off!), so we match inside of the {}s to get json only
try:
data = json.loads(re.search('(\{.+\})', phantom_output).group(1))
except Exception as e:
self.sql_driver.log_error(uri, "Could Not Load JSON: %s" % e)
return 'Could Not Load JSON'
# we need to parse the domain to determine if requests are local or 3rd party
# we need pubsuffix and tld for later analysis so store them now
origin_domain_pubsuffix_tld = self.uri_parser.get_domain_pubsuffix_tld(uri)
origin_domain = origin_domain_pubsuffix_tld[0]
origin_pubsuffix = origin_domain_pubsuffix_tld[1]
origin_tld = origin_domain_pubsuffix_tld[2]
if re.match('^Exception.+', origin_domain):
self.sql_driver.log_error(uri, 'Could not parse TLD for %s' % uri)
return 'Could not parse TLD for %s' % uri
page_domain_id = self.sql_driver.add_domain(origin_domain, origin_pubsuffix, origin_tld)
# newFollowLogger is giving us the follow JSON data:
# note source is null for now, but store anyway
# big_out = {
# final_uri: final_uri,
# title: page.title,
# meta_desc : meta_desc,
# requested_uris: JSON.stringify(requested_uris),
# received_uris: JSON.stringify(received_uris),
# cookies: phantom.cookies,
# source: 'NULL',
# };
# we are now storing uris with and without args and saving args
# we must unquote to uri to get back to original state so can parse
start_uri_original = urllib.parse.unquote(uri)
try:
start_uri_no_args = re.search('^(.+?)\?.+$', start_uri_original).group(1) # start uri no args
except:
start_uri_no_args = uri
try:
start_uri_args = re.search('^.+(\?.+)$', start_uri_original).group(1) # start uri args
except:
start_uri_args = 'NULL'
# same for the final uri (this is where we are after potential redirects)
final_uri = re.sub('\"', '', json.dumps(data["final_uri"]))
final_uri_original = urllib.parse.unquote(final_uri)
try:
final_uri_no_args = re.search('^(.+?)\?.+$', final_uri_original).group(1) # start uri no args
except:
final_uri_no_args = final_uri
try:
final_uri_args = re.search('^.+(\?.+)$', final_uri_original).group(1) # start uri args
except:
final_uri_args = 'NULL'
# add page
# json.dumps to make sure strings go out ok for db
page_id = self.sql_driver.add_page(
str(re.sub('\"', '', json.dumps(data["title"]))),
str(re.sub('\"', '', json.dumps(data["meta_desc"]))),
uri,
start_uri_no_args,
start_uri_args,
final_uri,
final_uri_no_args,
final_uri_args,
str(re.sub('\"', '', json.dumps(data["source"]))),
str(re.sub('\"', '', json.dumps(data["requested_uris"]))),
str(re.sub('\"', '', json.dumps(data["received_uris"]))),
page_domain_id)
for cookie in data["cookies"]:
# store external cookies, uri_parser fails on non-http, we should fix this
# right now a lame hack is to prepend http://
cookie_domain_pubsuffix_tld = self.uri_parser.get_domain_pubsuffix_tld("http://"+cookie["domain"])
cookie_domain = cookie_domain_pubsuffix_tld[0]
cookie_pubsuffix = cookie_domain_pubsuffix_tld[1]
cookie_tld = cookie_domain_pubsuffix_tld[2]
# something went wrong, but carry on...
if re.match('^Exception.+', cookie_domain):
self.sql_driver.log_error(uri, 'Error parsing cookie: '+cookie_domain)
continue
# this is a 3party cookie
if origin_domain != cookie_domain:
cookie_domain_id = self.sql_driver.add_domain(cookie_domain, cookie_pubsuffix, cookie_tld)
# name and domain are required, so if they fail we just continue
try: name = cookie["name"]
except: continue
try: domain = cookie_domain
except: continue
# these are optional, keep going with "N/A" vals
try: secure = cookie["secure"]
except: secure = "N/A"
try: path = cookie["path"]
except: path = "N/A"
try: expires = cookie["expires"]
except: expires = "N/A"
try: httponly = cookie["httponly"]
except: httponly = "N/A"
try: expiry = cookie["expiry"]
except: expiry = "N/A"
try: value = cookie["value"]
except: value = "N/A"
cookie_id = self.sql_driver.add_cookie( name, secure, path, domain,
expires, httponly, expiry, value,
cookie_domain_id)
self.sql_driver.add_cookie_to_page(cookie_id, page_id)
for request in data["requested_uris"]:
# if the request starts with "data" we can't parse tld anyway, so skip
if re.match('^(data|about|chrome).+', request):
continue
# get domain, pubsuffix, and tld from request
requested_domain_pubsuffix_tld = self.uri_parser.get_domain_pubsuffix_tld(request)
requested_domain = requested_domain_pubsuffix_tld[0]
requested_pubsuffix = requested_domain_pubsuffix_tld[1]
requested_tld = requested_domain_pubsuffix_tld[2]
# see if we got back what we requested, if not a few things may have happened
# * malformed uri
# * resource is gone or never existed
# * network latency (ie it didn't arrive in window specified)
# * we could be behind a firewall or censorship mechanism (eg gfw, golden shield)
# * our IP is blacklisted b/c we are totally a bot X-D
# the point being, interpret this value with an open mind
if request in data['received_uris']:
recieved = '1'
else:
recieved = '0'
# catch exceptions
if re.match('^Exception.+', requested_domain):
self.sql_driver.log_error(uri, 'Error parsing element request: '+request)
continue
# store new elements
if origin_domain != requested_domain:
full_uri = request
try:
element_uri = re.search('^(.+?)\?.+$', full_uri).group(1) # start uri no args
except:
element_uri = full_uri
# attempt to parse off the extension
try:
element_extension = re.search('\.([0-9A-Za-z]+)$', element_uri).group(1).lower()
except:
element_extension = "NULL"
# figure out what type of element it is
if element_extension == 'png' or element_extension == 'jpg' or element_extension == 'jpgx' or element_extension == 'jpeg' or element_extension == 'gif' or element_extension == 'svg' or element_extension == 'bmp' or element_extension == 'tif' or element_extension == 'tiff' or element_extension == 'webp' or element_extension == 'srf':
element_type = 'image'
elif element_extension == 'js' or element_extension == 'javascript':
element_type = 'javascript'
elif element_extension == 'json' or element_extension == 'jsonp' or element_extension == 'xml':
element_type = 'data_structured'
elif element_extension == 'css':
element_type = 'style_sheet'
elif element_extension == 'woff' or element_extension == 'ttf' or element_extension == 'otf':
element_type = 'font'
elif element_extension == 'htm' or element_extension == 'html' or element_extension == 'shtml':
element_type = 'page_static'
elif element_extension == 'php' or element_extension == 'asp' or element_extension == 'jsp' or element_extension == 'aspx' or element_extension == 'ashx' or element_extension == 'pl' or element_extension == 'cgi' or element_extension == 'fcgi':
element_type = 'page_dynamic'
elif element_extension == 'swf' or element_extension == 'fla':
element_type = 'Shockwave Flash'
elif element_extension == 'NULL':
element_type = 'NULL'
else:
element_type = 'unknown'
try:
args = re.search('^.+(\?.+)$', full_uri).group(1) # start uri args
except:
args = 'NULL'
element_domain_id = self.sql_driver.add_domain(requested_domain, requested_pubsuffix, requested_tld)
element_id = self.sql_driver.add_element("NULL", full_uri, element_uri, recieved, element_extension, element_type, args, element_domain_id)
self.sql_driver.add_element_to_page(element_id, page_id)
return 'Successfully Added to DB'
# end report()
def close(self):
# close mysql connections
self.uri_parser.close()
self.sql_driver.close()
return
def __init__(self, dbname):
self.uri_parser = ParseURI()
self.sql_driver = MySQLDriver(dbname)
class ParseURI:
def __init__(self):
# load up the tld list now as only hit it once this way
self.pubsuffix_list = self.get_pubsuffix_list()
# this is to speed up tlds lookup with hash table
# we can share among all runs over time
self.sql_driver = MySQLDriver()
# check if sub_domain_tld exists, if not create one
# this should only really happen once
#
# bug?
# when using pool this can be done several times in tandem
# resulting in some lost entries - e.g. two threads see db not exist the first
# one to break the tie will create a new blank db and add a new entry
# it is possible the slightly slower second thread will also create a blank db
# erasing the entry of the first thread, however, after the first round the db
# should not be re-created again
#
# it should be emphasized this db is ONLY used to speed up parsing the tld
# (partially because my regex method is slow and should be refactored!)
# the data is this db is NOT used for analysis - thus deleting a few records
# will have a minor impact on speed for the first few entries, and then be
# wholly irrelevant
#
# still...it irks me that this entire class could be smarter, so I will
# fix it another day
if self.sql_driver.check_db_exist('sub_domain_tld') == False:
self.sql_driver.create_sub_domain_tld_db()
else:
self.sql_driver.db_switch('sub_domain_tld')
# end __init__
def get_pubsuffix_list(self):
# get the file from the local dir
pubsuffix_raw_list = open(
os.path.join(
os.path.dirname(__file__),
'./resources/pubsuffix/patchedPublicSuffixList-20150514.txt'),
'r')
pubsuffix_list = []
for line in pubsuffix_raw_list:
# the last part of the list is random shit we don't care about, so stop reading
if re.match("^// ===BEGIN PRIVATE DOMAINS===", line): break
# skip lines that are comments or blank, add others to list
# also remove leading ., !, and * as it f***s up regex later
if not re.match("^//.+$|^$", line):
pubsuffix_list.append(re.sub('^[\!\*]\.?', '', line.strip()))
# we sort long->short so we can match deeper TLD first (ie, ac.uk *before* .uk)
pubsuffix_list.sort(key=len, reverse=True)
# to speed things up we move the most common TLD to the front of the line
# that said, if it's not one of these we take a MASSIVE performance hit
popular_pubsuffixs = [
'gov', 'co.uk', 'mil', 'org', 'net', 'edu', 'com'
]
for popular_pubsuffix in popular_pubsuffixs:
pubsuffix_list.remove(popular_pubsuffix)
pubsuffix_list.insert(0, popular_pubsuffix)
return pubsuffix_list
# get_pubsuffix_list
def get_domain_pubsuffix_tld(self, uri):
# pull out the first chunk of the domain, possibly including subdomains
# if this fails, the domain is f****d up (or not https? at least), kill
# only handles if the domain is followed by $, ?, \, or / - other shit will break...
# adding to handle with port at end, ie blah.com:8080, so rule is '|\:[0-9].+
# adding \.? at the top to handle leading '.'
try:
sub_domain = re.search(
'^https?:\/\/\.?(.*?)(\:[0-9].+)?($|\/|\?|\\\\|=)',
uri).group(1)
except:
return ('Exception: Unable to parse: ' + uri[:50],
'Exception: Unable to parse: ' + uri[:50],
'Exception: Unable to parse: ' + uri[:50])
# strip off leading or trailing '.', this is f*****g things up
sub_domain = re.sub('(^\.|\.$)', '', sub_domain)
# see if it is an ip address, if so return it
# this is only ipv4 pattern though, maybe should thinking about ipv6?
try:
re.search('(^\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}$)',
sub_domain).group(1)
return (sub_domain, 'None', 'None')
except:
pass
# first, we will see if it is in the db already
record = self.sql_driver.sub_domain_exists(sub_domain)
if record:
return record
# the pubsuffix list is large -> small,
# so the first match is the one we want
# after we find it, break out and continue
for pubsuffix_try in self.pubsuffix_list:
if re.match(".+\." + pubsuffix_try + "$", sub_domain):
pubsuffix = pubsuffix_try
break
# if we didn't find the pubsuffix we fail
try:
pubsuffix
except:
return ('Exception: Unable to parse: ' + uri[:50],
'Exception: Unable to parse: ' + uri[:50],
'Exception: Unable to parse: ' + uri[:50])
# if we have sub.domain.tld_match, we just want domain.tld_match
# there is no reason this should fail if we get this far, but try/except to be safe
try:
domain = re.search('(.*\.)?\.?(.*\.' + pubsuffix + ')$',
sub_domain).group(2)
except:
return ('Exception: Unable to parse: ' + uri[:50],
'Exception: Unable to parse: ' + uri[:50],
'Exception: Unable to parse: ' + uri[:50])
# grab the tld off of the pubsuffix
# if regex fails the tld and pubsuffix are the same
try:
tld = re.search('\.([0-9A-Za-z]+)$', pubsuffix).group(1).lower()
except:
tld = pubsuffix
self.sql_driver.add_sub_domain_pubsuffix_tld(sub_domain, domain,
pubsuffix, tld)
return (domain, pubsuffix, tld)
#end get_domain_pubsuffix_tld
def close(self):
# close mysql connection
self.sql_driver.close()
return
# end close
#end ParseURI
def __init__(self,
db_engine,
db_name,
num_tlds,
num_results,
tracker_threshold=None,
flush_owner_db=True):
"""
This performs a few start-up tasks:
- sets up some useful global variables
- makes sure we have a directory to store the reports
- flushes the existing domain_owner mappings (this can be disabled)
- if we want to do per-tld reports, figures out the most common
- if we want to filter against a given tracker threshold, sets it
up here (see documentation below for tracker threshold)
"""
# set various global vars
self.db_engine = db_engine
self.db_name = db_name
self.num_tlds = num_tlds
self.top_tlds = []
self.num_results = num_results
self.tracker_threshold = tracker_threshold
self.start_time = datetime.now()
# number of decimal places to round to in reports
self.num_decimals = 2
# set up global db connection
if self.db_engine == 'mysql':
from webxray.MySQLDriver import MySQLDriver
self.sql_driver = MySQLDriver(self.db_name)
elif self.db_engine == 'sqlite':
from webxray.SQLiteDriver import SQLiteDriver
self.sql_driver = SQLiteDriver(self.db_name)
elif db_engine == 'postgres':
from webxray.PostgreSQLDriver import PostgreSQLDriver
self.sql_driver = PostgreSQLDriver(self.db_name)
else:
print('INVALID DB ENGINE FOR %s, QUITTING!' % db_engine)
exit()
# this is reused often, do it once to save time
self.get_pages_ok_count = self.sql_driver.get_pages_ok_count()
print('\t=============================')
print('\t Checking Output Directories ')
print('\t=============================')
self.setup_report_dir()
print('\t============================')
print('\t Patching Domain Owner Data ')
print('\t============================')
if flush_owner_db:
# update the domains to their owners in the db, can be overridden
# by changing flush_owner_db to false
self.patch_domain_owners()
else:
print('\t\t\tSkipping')
# this is used in various places to get owner information
self.domain_owners = self.get_domain_owner_dict()
# if we want to get sub-reports for the most frequent tlds we find
# them here
if self.num_tlds:
print('\t=====================')
print('\t Getting top %s tlds' % self.num_tlds)
print('\t=====================')
print('\t\tProcessing...', end='', flush=True)
self.top_tlds = self.get_top_tlds(self.num_tlds)
print('done!')
print('\t\tThe top tlds are:')
for (tld, pages) in self.top_tlds:
if tld: print('\t\t |- %s (%s)' % (tld, pages))
else:
# othewise we push in a single empty entry
self.top_tlds.append((None, self.get_pages_ok_count))
# SPECIAL FEATURE FOR EXPERTS: tracker domain filter
#
# you can set a threshold of the number of sites a given 3p domain
# is connected to - domains connecting to many sites may correlate those visits
# so we call these 'tracker domains'
#
# the 'tracker_threshold' variable set above controls the filtering level
#
# on large set of sites (e.g. >10k) this works well but on small samples
# (e.g. <500) it doesn't work as well as known tracker domains may only
# appear on a single site
#
# this is off by default and unless you understand what you are doing
# don't use this...but because you are reading the source code for an otherwise
# undocumented feature you are probably competent to use it ;-)
#
# longer-term we may want to train off a bigger corpus to find tracker domains and
# have them prepackaged
#
# use at your own risk!
if tracker_threshold:
print('\t===================================================')
print('\t Getting tracker domains with threshold level of %s' %
self.tracker_threshold)
print('\t===================================================')
print('\t\tProcessing...', end='', flush=True)
self.tracker_domains = self.get_tracker_domains(
self.tracker_threshold)
print('done!')
else:
# set to None so various downstream operations get skipped
self.tracker_domains = None
def store(self, url, browser_output, store_source=False, store_1p=True, get_file_hashes=False, hash_3p_only=False):
"""
this is the primary function of this class,
it takes the url of the given page and the request and cookie data generated
by the browser
data is cleaned up with some minor analysis (eg file types) and stored
for later in-depth analysis.
there is an option to store first party requests as well as third, turned on by default
to save disk space turn off store_1p
there is also an option to get file hashes, this introduces serious overhead
and is turned off by default
"""
# open up a sql connection
if self.db_engine == 'mysql':
from webxray.MySQLDriver import MySQLDriver
sql_driver = MySQLDriver(self.db_name)
elif self.db_engine == 'sqlite':
from webxray.SQLiteDriver import SQLiteDriver
sql_driver = SQLiteDriver(self.db_name)
elif self.db_engine == 'postgres':
from webxray.PostgreSQLDriver import PostgreSQLDriver
sql_driver = PostgreSQLDriver(self.db_name)
else:
print('INVALED DB ENGINE FOR %s, QUITTING!' % db_engine)
exit()
# get the ip, fqdn, domain, pubsuffix, and tld
# we need the domain to figure out if cookies/elements are third-party
origin_ip_fqdn_domain_pubsuffix_tld = self.url_parser.get_ip_fqdn_domain_pubsuffix_tld(url)
# if we can't get page domain info we fail gracefully
if origin_ip_fqdn_domain_pubsuffix_tld is None:
sql_driver.log_error(url, 'Could not parse TLD for %s' % url)
return False
origin_ip = origin_ip_fqdn_domain_pubsuffix_tld[0]
origin_fqdn = origin_ip_fqdn_domain_pubsuffix_tld[1]
origin_domain = origin_ip_fqdn_domain_pubsuffix_tld[2]
origin_pubsuffix = origin_ip_fqdn_domain_pubsuffix_tld[3]
origin_tld = origin_ip_fqdn_domain_pubsuffix_tld[4]
# sql_driver.add_domain both stores the new domain and returns its db row id
# if it is already in db just return the existing id
page_domain_id = sql_driver.add_domain(origin_ip, origin_fqdn, origin_domain, origin_pubsuffix, origin_tld)
# figure out the privacy policy url and text, starts null
priv_policy_url = None
priv_policy_url_text = None
# read in our list of privacy link terms from the json file in webxray/resources/policyxray
privacy_policy_term_list = self.utilities.get_privacy_policy_term_list()
# we reverse links return from browser to check footer links first as that is where policy links tend to be
all_links = browser_output['all_links']
all_links.reverse()
# if we have links search for privacy policy
if len(all_links) > 0:
# links are tuple
for link_text,link_url in all_links:
# makes sure we have text, skip links without
if link_text:
# need lower for string matching
link_text = link_text.lower().strip()
# not a link we can use
if 'javascript' in link_text: continue
# see if the link_text is in our term list
if link_text in privacy_policy_term_list:
# if the link_url is relative this will convert to absolute
priv_policy_url = self.utilities.get_absolute_url_from_page_link(url,link_url)
priv_policy_url_text = link_text
break
# if the final page is https (often after a redirect), mark it appropriately
if browser_output['final_url'][:5] == 'https':
page_is_ssl = True
else:
page_is_ssl = False
if store_source:
# handles issue where postgres will crash on inserting null character
source = browser_output['source'].replace('\x00',' ')
else:
source = None
# add page
page_id = sql_driver.add_page(
browser_output['browser_type'],
browser_output['browser_version'],
browser_output['browser_wait'],
browser_output['title'],
browser_output['meta_desc'],
url,
browser_output['final_url'],
priv_policy_url,
priv_policy_url_text,
page_is_ssl,
source,
browser_output['load_time'],
page_domain_id
)
# store cookies
for cookie in browser_output['cookies']:
# get the ip, fqdn, domain, pubsuffix, and tld
# we need the domain to figure out if cookies/elements are third-party
# note:
# url_parser fails on non-http, we should fix this, right now a lame hack is to prepend http://
cookie_ip_fqdn_domain_pubsuffix_tld = self.url_parser.get_ip_fqdn_domain_pubsuffix_tld('http://'+cookie['domain'])
# something went wrong, log and fail gracefully
if cookie_ip_fqdn_domain_pubsuffix_tld is None:
sql_driver.log_error(url, 'Error parsing cookie with domain: '+cookie['domain'])
continue
# otherwise, everything went fine
cookie_ip = cookie_ip_fqdn_domain_pubsuffix_tld[0]
cookie_fqdn = cookie_ip_fqdn_domain_pubsuffix_tld[1]
cookie_domain = cookie_ip_fqdn_domain_pubsuffix_tld[2]
cookie_pubsuffix = cookie_ip_fqdn_domain_pubsuffix_tld[3]
cookie_tld = cookie_ip_fqdn_domain_pubsuffix_tld[4]
# mark third-party cookies
if origin_domain != cookie_domain:
is_3p_cookie = True
else:
is_3p_cookie = False
# this is a first party cookie, see if we want to store it
if is_3p_cookie is False and store_1p is False:
continue
# sql_driver.add_domain both stores the new domain and returns its id
cookie_domain_id = sql_driver.add_domain(cookie_ip, cookie_fqdn, cookie_domain, cookie_pubsuffix, cookie_tld)
# name and domain are required, so if they fail we just continue
try: name = cookie['name']
except: continue
try: domain = cookie_domain
except: continue
# these are optional, fill with null values if fail
try: secure = cookie['secure']
except: secure = None
try: path = cookie['path']
except: path = None
try: httponly = cookie['httponly']
except: httponly = None
try: expiry = cookie['expiry']
except: expiry = None
try: value = cookie['value']
except: value = None
# all done with this cookie
sql_driver.add_cookie(
page_id,
name, secure, path, domain,
httponly, expiry, value,
is_3p_cookie, cookie_domain_id
)
# process requests now
for request in browser_output['processed_requests']:
# if the request starts with the following we can't parse anyway, so skip
if re.match('^(data|about|chrome|blob).+', request):
continue
# get the ip, fqdn, domain, pubsuffix, and tld
# we need the domain to figure out if cookies/elements are third-party
element_ip_fqdn_domain_pubsuffix_tld = self.url_parser.get_ip_fqdn_domain_pubsuffix_tld(request)
# problem with this request, log and fail gracefully
if element_ip_fqdn_domain_pubsuffix_tld is None:
sql_driver.log_error(url, 'Error parsing element request: '+request)
continue
element_ip = element_ip_fqdn_domain_pubsuffix_tld[0]
element_fqdn = element_ip_fqdn_domain_pubsuffix_tld[1]
element_domain = element_ip_fqdn_domain_pubsuffix_tld[2]
element_pubsuffix = element_ip_fqdn_domain_pubsuffix_tld[3]
element_tld = element_ip_fqdn_domain_pubsuffix_tld[4]
# sql_driver.add_domain both stores the new domain and returns its db row id
element_domain_id = sql_driver.add_domain(element_ip, element_fqdn, element_domain, element_pubsuffix, element_tld)
# mark third-party elements based on domain
if origin_domain != element_domain:
is_3p_element = True
else:
is_3p_element = False
# if we are not storing 1p elements continue
if is_3p_element is False and store_1p is False:
continue
if request[:5] == 'https' or request[:3] == 'wss':
element_is_ssl = True
else:
element_is_ssl = False
try:
received = browser_output['processed_requests'][request]['received']
except:
received = None
# get domain of referer and determine if page leaked by referer
try:
referer = browser_output['processed_requests'][request]['referer']
except:
referer = None
if referer and len(referer) != 0:
referer_ip_fqdn_domain_pubsuffix_tld = self.url_parser.get_ip_fqdn_domain_pubsuffix_tld(referer)
if referer_ip_fqdn_domain_pubsuffix_tld:
if referer_ip_fqdn_domain_pubsuffix_tld[2] == origin_domain:
page_domain_in_referer = True
else:
page_domain_in_referer = False
else:
page_domain_in_referer = None
sql_driver.log_error(url, 'Error parsing referer header: '+referer)
else:
page_domain_in_referer = None
try:
start_time_offset = browser_output['processed_requests'][request]['start_time_offset']
except:
start_time_offset = None
try:
load_time = browser_output['processed_requests'][request]['load_time']
except:
load_time = None
try:
status = browser_output['processed_requests'][request]['status']
except:
status = None
try:
status_text = browser_output['processed_requests'][request]['status_text']
except:
status_text = None
try:
content_type = browser_output['processed_requests'][request]['content_type']
except:
content_type = None
try:
body_size = browser_output['processed_requests'][request]['body_size']
except:
body_size = None
try:
request_headers = str(browser_output['processed_requests'][request]['request_headers'])
except:
request_headers = None
try:
response_headers = str(browser_output['processed_requests'][request]['response_headers'])
except:
response_headers = None
# consider anything before the "?" to be the element_url
try:
element_url = re.search('^(.+?)\?.+$', request).group(1)
except:
element_url = request
# consider anything after the "?" to be the args
try:
element_args = re.search('^.+(\?.+)$', request).group(1) # start url args
except:
element_args = None
# attempt to parse off the extension
try:
element_extension = re.search('\.([0-9A-Za-z]+)$', element_url).group(1).lower()
except:
element_extension = None
# lists of common extensions, can be expanded
image_extensions = ['png', 'jpg', 'jpgx', 'jpeg', 'gif', 'svg', 'bmp', 'tif', 'tiff', 'webp', 'srf']
script_extensions = ['js', 'javascript']
data_extensions = ['json', 'jsonp', 'xml']
font_extentions = ['woff', 'ttf', 'otf']
static_extentions = ['html', 'htm', 'shtml']
dynamic_extentions = ['php', 'asp', 'jsp', 'aspx', 'ashx', 'pl', 'cgi', 'fcgi']
# figure out what type of element it is
if element_extension in image_extensions:
element_type = 'image'
elif element_extension in script_extensions:
element_type = 'javascript'
elif element_extension in data_extensions:
element_type = 'data_structured'
elif element_extension == 'css':
element_type = 'style_sheet'
elif element_extension in font_extentions:
element_type = 'font'
elif element_extension in static_extentions:
element_type = 'page_static'
elif element_extension == dynamic_extentions:
element_type = 'page_dynamic'
elif element_extension == 'swf' or element_extension == 'fla':
element_type = 'Shockwave Flash'
else:
element_type = None
# file hashing has non-trivial overhead and off by default
#
# what this does is uses the same ua/referer as the actual request
# so we are just replaying the last one to get similar response
# note that we aren't sending the same cookies so that could be an issue
# otherwise it is equivalent to a page refresh in theory
# option to hash only 3p elements observed here
if (get_file_hashes and hash_3p_only and is_3p_element) or (get_file_hashes and hash_3p_only == False):
replay_element_request = urllib.request.Request(
request,
headers = {
'User-Agent' : browser_output['processed_requests'][request]['user_agent'],
'Referer' : referer,
'Accept' : '*/*'
}
)
try:
file_md5 = hashlib.md5(urllib.request.urlopen(replay_element_request,timeout=10).read()).hexdigest()
except:
file_md5 = None
else:
file_md5 = None
# final tasks is to truncate the request if it is
# over 2k characters as it is likely
# binary data and may cause problems inserting
# into TEXT fields in database
#
# TODO:
# better handle binary data in general
if len(request) >= 2000: request = request[:2000]
if len(element_url) >= 2000: element_url = element_url[:2000]
# store request
sql_driver.add_element(
page_id,
request, element_url,
is_3p_element, element_is_ssl,
received,
referer,
page_domain_in_referer,
start_time_offset,
load_time,
status,
status_text,
content_type,
body_size,
request_headers,
response_headers,
file_md5,
element_extension,
element_type,
element_args,
element_domain_id
)
# close db connection
sql_driver.close()
return True
def run(self, pool_size):
"""
this function manages the parallel processing of the url list using the python Pool class
the function first reads the list of urls out of the page_lists directory, cleans it
for known issues (eg common binary files), and issues with idna encoding (tricky!)
then the page list is mapped to the process_url function and executed in parallell
pool_size is defined in the run_webxray.py file, see details there
"""
# the list of url MUST be in the page_lists directory!
try:
url_list = open(os.path.dirname(os.path.abspath(__file__)) + '/../page_lists/' + self.pages_file_name, 'r', encoding='utf-8')
except:
print('File "%s" does not exist, file must be in ./page_lists directory. Exiting.' % self.pages_file_name)
exit()
# set up sql connection used to determine if items are already in the db
if self.db_engine == 'mysql':
from webxray.MySQLDriver import MySQLDriver
sql_driver = MySQLDriver(self.db_name)
elif self.db_engine == 'postgres':
from webxray.PostgreSQLDriver import PostgreSQLDriver
sql_driver = PostgreSQLDriver(self.db_name)
elif self.db_engine == 'sqlite':
from webxray.SQLiteDriver import SQLiteDriver
sql_driver = SQLiteDriver(self.db_name)
# this list gets mapped to the Pool, very important!
urls_to_process = set()
# simple counter used solely for updates to CLI
count = 0
print('\t------------------------')
print('\t Building List of Pages ')
print('\t------------------------')
for url in url_list:
# skip lines that are comments
if "#" in url[0]: continue
count += 1
# only do lines starting with https?://
if not (re.match('^https?://.+', url)):
print("\t\t%s | %-50s Not a valid address, Skipping." % (count, url[:50]))
continue
# non-ascii domains will crash phantomjs, so we need to convert them to
# idna/ascii/utf-8
# this requires splitting apart the url, converting the domain to idna,
# and pasting it all back together
split_url = urlsplit(url.strip())
idna_fixed_netloc = split_url.netloc.encode('idna').decode('utf-8')
url = urlunsplit((split_url.scheme,idna_fixed_netloc,split_url.path,split_url.query,split_url.fragment))
# if it is a m$ office or other doc, skip
if re.match('.+(pdf|ppt|pptx|doc|docx|txt|rtf|xls|xlsx)$', url):
print("\t\t%s | %-50s Not an HTML document, Skipping." % (count, url[:50]))
continue
# skip if in db already unless we are doing a timeseries
if self.allow_timeseries == False:
if sql_driver.page_exists(url):
print("\t\t%s | %-50s Exists in DB, Skipping." % (count, url[:50]))
continue
# only add if not in list already
if url not in urls_to_process:
print("\t\t%s | %-50s Adding." % (count, url[:50]))
urls_to_process.add(url)
else:
print("\t\t%s | %-50s Already queued, Skipping." % (count, url[:50]))
# close the db connection
sql_driver.close()
print('\t----------------------------------')
print('\t%s addresses will now be webXray\'d' % len(urls_to_process))
print('\t\tBrowser(s) are %s' % self.browser_types)
print('\t\tBrowser wait time is %s seconds' % self.browser_wait)
print('\t\t...you can go take a walk. ;-)')
print('\t----------------------------------')
# for macOS (darwin) we must specify start method as 'forkserver'
# this is essentially voodoo to ward off evil spirits which
# appear when large pool sizes are used on macOS
# get_start_method must be set to 'allow_none', otherwise upon
# checking the method it gets set (!) - and if we then get/set again
# we get an error
if sys.platform == 'darwin' and multiprocessing.get_start_method(allow_none=True) != 'forkserver':
multiprocessing.set_start_method('forkserver')
myPool = multiprocessing.Pool(pool_size)
myPool.map(self.process_url, urls_to_process)
# FYI
self.print_runtime()
def process_url(self, url):
"""
this function takes a specified url, loads it in the browser (currently phantomjs)
and returns json-formatted output with relevant request data, etc.
the output_store class then puts this data in the db for later analysis
"""
# set up sql connection used to log errors and do timeseries checks
if self.db_engine == 'mysql':
from webxray.MySQLDriver import MySQLDriver
sql_driver = MySQLDriver(self.db_name)
elif self.db_engine == 'postgres':
from webxray.PostgreSQLDriver import PostgreSQLDriver
sql_driver = PostgreSQLDriver(self.db_name)
elif self.db_engine == 'sqlite':
from webxray.SQLiteDriver import SQLiteDriver
sql_driver = SQLiteDriver(self.db_name)
# output store does the heavy lifting of analyzing browser output and storing to db
output_store = OutputStore(self.db_engine, self.db_name)
# support for loading same page with multiple browsers - purposefully undocumented
for browser_type in self.browser_types:
# import and set up specified browser driver
# note we need to set up a new browser each time to
# get a fresh profile
if browser_type == 'phantomjs':
browser_driver = PhantomDriver()
elif browser_type == 'chrome':
browser_driver = ChromeDriver(ua=self.chrome_ua)
# support for timeseries collections - purposefully undocumented
if self.allow_timeseries:
page_last_accessed_browser_type = sql_driver.get_page_last_accessed_by_browser_type(url,browser_type)
if page_last_accessed_browser_type:
time_diff = datetime.now()-page_last_accessed_browser_type[0]
if time_diff < timedelta(minutes=self.interval_minutes) and page_last_accessed_browser_type[1] == browser_type:
print("\t\t%-50s Scanned too recently with %s" % (url[:50], browser_type))
continue
# attempt to load the page, fail gracefully
try:
browser_output = browser_driver.get_webxray_scan_data(url, self.browser_wait)
except:
print('\t\t%-50s Browser %s Did Not Return' % (url[:50], browser_type))
sql_driver.log_error(url, 'Unable to load page')
sql_driver.close()
return
# if there was a problem browser_output will be None
if browser_output == None:
print('\t\t%-50s Browser %s Did Not Return' % (url[:50], browser_type))
sql_driver.log_error(url, 'Unable to load page')
sql_driver.close()
return
# attempt to store the output
if output_store.store(url, browser_output):
print('\t\t%-50s Success with %s' % (url[:50],browser_type))
else:
print('\t\t%-50s Fail with %s' % (url[:50],browser_type))
sql_driver.log_error(url, 'Unable to load page')
sql_driver.close()
return
def run(self, pool_size):
"""
this function manages the parallel processing of the url list using the python Pool class
the function first reads the list of urls out of the page_lists directory, cleans it
for known issues (eg common binary files), and issues with idna encoding (tricky!)
then the page list is mapped to the process_url function and executed in parallell
pool_size is defined in the run_webxray.py file, see details there
"""
# the list of url MUST be in the page_lists directory!
try:
url_list = open(os.path.dirname(os.path.abspath(__file__)) + '/../page_lists/' + self.pages_file_name, 'r')
except:
print('File "%s" does not exist, file must be in ./page_lists directory. Exiting.' % self.pages_file_name)
exit()
# set up sql connection used to determine if items are already in the db
if self.db_engine == 'mysql':
from webxray.MySQLDriver import MySQLDriver
sql_driver = MySQLDriver(self.db_name)
elif self.db_engine == 'postgres':
from webxray.PostgreSQLDriver import PostgreSQLDriver
sql_driver = PostgreSQLDriver(self.db_name)
elif self.db_engine == 'sqlite':
from webxray.SQLiteDriver import SQLiteDriver
sql_driver = SQLiteDriver(self.db_name)
# this list gets mapped to the Pool, very important!
urls_to_process = set()
# simple counter used solely for updates to CLI
count = 0
print('\t------------------------')
print('\t Building List of Pages ')
print('\t------------------------')
for url in url_list:
# skip lines that are comments
if "#" in url[0]: continue
count += 1
# only do lines starting with https?://
if not (re.match('^https?://.+', url)):
print("\t\t%s | %-50s Not a valid address, Skipping." % (count, url[:50]))
continue
# non-ascii domains will crash phantomjs, so we need to convert them to
# idna/ascii/utf-8
# this requires splitting apart the url, converting the domain to idna,
# and pasting it all back together
split_url = urlsplit(url.strip())
idna_fixed_netloc = split_url.netloc.encode('idna').decode('utf-8')
url = urlunsplit((split_url.scheme,idna_fixed_netloc,split_url.path,split_url.query,split_url.fragment))
# if it is a m$ office or other doc, skip
if re.match('.+(pdf|ppt|pptx|doc|docx|txt|rtf|xls|xlsx)$', url):
print("\t\t%s | %-50s Not an HTML document, Skipping." % (count, url[:50]))
continue
# skip if in db already unless we are doing a timeseries
if self.allow_timeseries == False:
if sql_driver.page_exists(url):
print("\t\t%s | %-50s Exists in DB, Skipping." % (count, url[:50]))
continue
# only add if not in list already
if url not in urls_to_process:
print("\t\t%s | %-50s Adding." % (count, url[:50]))
urls_to_process.add(url)
else:
print("\t\t%s | %-50s Already queued, Skipping." % (count, url[:50]))
# close the db connection
sql_driver.close()
print('\t----------------------------------')
print('\t%s addresses will now be webXray\'d' % len(urls_to_process))
print('\t\tBrowser(s) are %s' % self.browser_types)
print('\t\tBrowser wait time is %s seconds' % self.browser_wait)
print('\t\t...you can go take a walk. ;-)')
print('\t----------------------------------')
# for macOS (darwin) we must specify start method as 'forkserver'
# this is essentially voodoo to ward off evil spirits which
# appear when large pool sizes are used on macOS
# get_start_method must be set to 'allow_none', otherwise upon
# checking the method it gets set (!) - and if we then get/set again
# we get an error
if sys.platform == 'darwin' and multiprocessing.get_start_method(allow_none=True) != 'forkserver':
multiprocessing.set_start_method('forkserver')
myPool = multiprocessing.Pool(pool_size)
myPool.map(self.process_url, urls_to_process)
# FYI
self.print_runtime()
class OutputStore:
def __init__(self, dbname):
self.uri_parser = ParseURI()
self.sql_driver = MySQLDriver(dbname)
# end init
def store(self, uri, phantom_output):
# parse out the json from our phantom_output
# sometimes phantom prints out errors before the json, (despite us turning
# it off!), so we match inside of the {}s to get json only
try:
data = json.loads(re.search('(\{.+\})', phantom_output).group(1))
except Exception as e:
self.sql_driver.log_error(uri, "Could Not Load JSON: %s" % e)
return 'Could Not Load JSON'
# we need to parse the domain to determine if requests are local or 3rd party
# we need pubsuffix and tld for later analysis so store them now
origin_domain_pubsuffix_tld = self.uri_parser.get_domain_pubsuffix_tld(
uri)
origin_domain = origin_domain_pubsuffix_tld[0]
origin_pubsuffix = origin_domain_pubsuffix_tld[1]
origin_tld = origin_domain_pubsuffix_tld[2]
if re.match('^Exception.+', origin_domain):
self.sql_driver.log_error(uri, 'Could not parse TLD for %s' % uri)
return 'Could not parse TLD for %s' % uri
page_domain_id = self.sql_driver.add_domain(origin_domain,
origin_pubsuffix,
origin_tld)
# newFollowLogger is giving us the follow JSON data:
# note source is null for now, but store anyway
# big_out = {
# final_uri: final_uri,
# title: page.title,
# meta_desc : meta_desc,
# requested_uris: JSON.stringify(requested_uris),
# received_uris: JSON.stringify(received_uris),
# cookies: phantom.cookies,
# source: 'NULL',
# };
# we are now storing uris with and without args and saving args
# we must unquote to uri to get back to original state so can parse
start_uri_original = urllib.parse.unquote(uri)
try:
start_uri_no_args = re.search('^(.+?)\?.+$',
start_uri_original).group(
1) # start uri no args
except:
start_uri_no_args = uri
try:
start_uri_args = re.search('^.+(\?.+)$', start_uri_original).group(
1) # start uri args
except:
start_uri_args = 'NULL'
# same for the final uri (this is where we are after potential redirects)
final_uri = re.sub('\"', '', json.dumps(data["final_uri"]))
final_uri_original = urllib.parse.unquote(final_uri)
try:
final_uri_no_args = re.search('^(.+?)\?.+$',
final_uri_original).group(
1) # start uri no args
except:
final_uri_no_args = final_uri
try:
final_uri_args = re.search('^.+(\?.+)$', final_uri_original).group(
1) # start uri args
except:
final_uri_args = 'NULL'
# add page
# json.dumps to make sure strings go out ok for db
page_id = self.sql_driver.add_page(
str(re.sub('\"', '', json.dumps(data["title"]))),
str(re.sub('\"', '',
json.dumps(data["meta_desc"]))), uri, start_uri_no_args,
start_uri_args, final_uri, final_uri_no_args, final_uri_args,
str(re.sub('\"', '', json.dumps(data["source"]))),
str(re.sub('\"', '', json.dumps(data["requested_uris"]))),
str(re.sub('\"', '',
json.dumps(data["received_uris"]))), page_domain_id)
for cookie in data["cookies"]:
# store external cookies, uri_parser fails on non-http, we should fix this
# right now a lame hack is to prepend http://
cookie_domain_pubsuffix_tld = self.uri_parser.get_domain_pubsuffix_tld(
"http://" + cookie["domain"])
cookie_domain = cookie_domain_pubsuffix_tld[0]
cookie_pubsuffix = cookie_domain_pubsuffix_tld[1]
cookie_tld = cookie_domain_pubsuffix_tld[2]
# something went wrong, but carry on...
if re.match('^Exception.+', cookie_domain):
self.sql_driver.log_error(
uri, 'Error parsing cookie: ' + cookie_domain)
continue
# this is a 3party cookie
if origin_domain != cookie_domain:
cookie_domain_id = self.sql_driver.add_domain(
cookie_domain, cookie_pubsuffix, cookie_tld)
# name and domain are required, so if they fail we just continue
try:
name = cookie["name"]
except:
continue
try:
domain = cookie_domain
except:
continue
# these are optional, keep going with "N/A" vals
try:
secure = cookie["secure"]
except:
secure = "N/A"
try:
path = cookie["path"]
except:
path = "N/A"
try:
expires = cookie["expires"]
except:
expires = "N/A"
try:
httponly = cookie["httponly"]
except:
httponly = "N/A"
try:
expiry = cookie["expiry"]
except:
expiry = "N/A"
try:
value = cookie["value"]
except:
value = "N/A"
cookie_id = self.sql_driver.add_cookie(name, secure, path,
domain, expires,
httponly, expiry, value,
cookie_domain_id)
self.sql_driver.add_cookie_to_page(cookie_id, page_id)
for request in data["requested_uris"]:
# if the request starts with "data" we can't parse tld anyway, so skip
if re.match('^(data|about|chrome).+', request):
continue
# get domain, pubsuffix, and tld from request
requested_domain_pubsuffix_tld = self.uri_parser.get_domain_pubsuffix_tld(
request)
requested_domain = requested_domain_pubsuffix_tld[0]
requested_pubsuffix = requested_domain_pubsuffix_tld[1]
requested_tld = requested_domain_pubsuffix_tld[2]
# see if we got back what we requested, if not a few things may have happened
# * malformed uri
# * resource is gone or never existed
# * network latency (ie it didn't arrive in window specified)
# * we could be behind a firewall or censorship mechanism (eg gfw, golden shield)
# * our IP is blacklisted b/c we are totally a bot X-D
# the point being, interpret this value with an open mind
if request in data['received_uris']:
recieved = '1'
else:
recieved = '0'
# catch exceptions
if re.match('^Exception.+', requested_domain):
self.sql_driver.log_error(
uri, 'Error parsing element request: ' + request)
continue
# store new elements
if origin_domain != requested_domain:
full_uri = request
try:
element_uri = re.search('^(.+?)\?.+$', full_uri).group(
1) # start uri no args
except:
element_uri = full_uri
# attempt to parse off the extension
try:
element_extension = re.search(
'\.([0-9A-Za-z]+)$', element_uri).group(1).lower()
except:
element_extension = "NULL"
# figure out what type of element it is
if element_extension == 'png' or element_extension == 'jpg' or element_extension == 'jpgx' or element_extension == 'jpeg' or element_extension == 'gif' or element_extension == 'svg' or element_extension == 'bmp' or element_extension == 'tif' or element_extension == 'tiff' or element_extension == 'webp' or element_extension == 'srf':
element_type = 'image'
elif element_extension == 'js' or element_extension == 'javascript':
element_type = 'javascript'
elif element_extension == 'json' or element_extension == 'jsonp' or element_extension == 'xml':
element_type = 'data_structured'
elif element_extension == 'css':
element_type = 'style_sheet'
elif element_extension == 'woff' or element_extension == 'ttf' or element_extension == 'otf':
element_type = 'font'
elif element_extension == 'htm' or element_extension == 'html' or element_extension == 'shtml':
element_type = 'page_static'
elif element_extension == 'php' or element_extension == 'asp' or element_extension == 'jsp' or element_extension == 'aspx' or element_extension == 'ashx' or element_extension == 'pl' or element_extension == 'cgi' or element_extension == 'fcgi':
element_type = 'page_dynamic'
elif element_extension == 'swf' or element_extension == 'fla':
element_type = 'Shockwave Flash'
elif element_extension == 'NULL':
element_type = 'NULL'
else:
element_type = 'unknown'
try:
args = re.search('^.+(\?.+)$',
full_uri).group(1) # start uri args
except:
args = 'NULL'
element_domain_id = self.sql_driver.add_domain(
requested_domain, requested_pubsuffix, requested_tld)
element_id = self.sql_driver.add_element(
"NULL", full_uri, element_uri, recieved, element_extension,
element_type, args, element_domain_id)
self.sql_driver.add_element_to_page(element_id, page_id)
return 'Successfully Added to DB'
# end report()
def close(self):
# close mysql connections
self.uri_parser.close()
self.sql_driver.close()
return
class Reporter:
def __init__(self, db_name, num_tlds, num_results, tracker_threshold = 0):
self.db_name = db_name
self.sql_driver = MySQLDriver(self.db_name)
self.num_tlds = num_tlds
self.num_results = num_results
self.tracker_threshold = tracker_threshold
self.startTime = datetime.now()
self.pages_ok_count = self.sql_driver.pages_ok_count()
print('\t=============================')
print('\t Checking Output Directories ')
print('\t=============================')
self.setup_report_dir()
print('\t===========================')
print('\t Patching DB with Org Data ')
print('\t===========================')
# update the domains to their owners
self.patch_org_data()
print('\t\tSuccess!')
# if num_tlds is 0 we don't care about sub-analysis on the tlds (which takes time)
# so we just push in the wildcard with the total page count
# note that this is the *default* behavior
if self.num_tlds == 0:
self.top_tlds = []
self.top_tlds.append(('*',self.pages_ok_count))
# otherwise, we *do* care about the tlds, so get them
else:
print('\t=====================')
print('\t Getting top %s tlds' % self.num_tlds)
print('\t=====================')
print('\t\tProcessing...')
self.top_tlds = self.get_top_tlds(self.num_tlds)
print('\t\tSuccess!')
print('\t\tThe top tlds are:')
for (tld, pages) in self.top_tlds:
print('\t\t |- %s (%s)' % (tld,pages))
# SPECIAL SAUCE, FOR EXPERTS: tracker domains!
#
# idea for this is you set a threshold of the number of sites a given domain
# is connected to - domains connecting to many sites may correlate those visits
# via referer strings etc, so we call these 'tracker domains'
#
# on a really large set of sites (e.g. 1M+) this works well but on small samples
# (e.g. 500) it doesn't work well at all as known tracker domains may only
# appear on a single site
#
# this is off by default and unless you understand what you are doing...
# DON'T USE THIS!
#
# longer-term we may want to train off a bigger corpus to find tracker domains and
# have them prepackaged
#
if tracker_threshold:
print('\t=========================')
print('\t Getting tracker domains ')
print('\t=========================')
print('\t\tProcessing...')
self.tracker_domains = self.get_tracker_domains(self.tracker_threshold)
print('\t\tSuccess!')
else:
# set to False so various downstream operations get skipped
self.tracker_domains = False
# end __init__
#########################
# HELPERS, GENERAL #
#########################
def setup_report_dir(self):
# create directory for where the reports go if not exist
if os.path.exists('./reports') == False:
print('\t\tMaking global reports directory at ./reports.')
os.makedirs('./reports')
# set global report_path, trim off the wbxr_ prefix
self.report_path = './reports/'+self.db_name[5:]
# set up subdir for this database analysis
if os.path.exists(self.report_path) == False:
print('\t\tMaking subdirectory for reports at %s' % self.report_path)
os.makedirs(self.report_path)
# just a notice
print('\t\tStoring output in %s' % self.report_path)
# setup_report_dir
# reuse this a lot
def write_csv(self, file_name, csv_rows):
full_file_path = self.report_path+'/'+file_name
file_out = open(full_file_path, 'w')
for row in csv_rows:
file_out.write(row)
file_out.close()
print('\t\t'+'*'*40)
print('\t\tOutput written to %s' % full_file_path)
# write_csv
# just fyi
def print_runtime(self):
print('~='*40)
print("Finished!")
print("Time to process: "+str(datetime.now()-self.startTime)+"\n")
print('-'*80)
# end print_runtime
# X-(
def fatal(self, msg):
print('FATAL ERROR: %s' % msg)
print('EXITING.')
exit()
# fatal
#############################
# HELPERS, DATABASE/INIT #
#############################
def patch_org_data(self):
# in order to analyze what entities receive user data, we need to update
# the database with domain ownership records we have store previously
#
# we first clear out what is in there in case the new data has changed
# perhaps make this optional, on big dbs takes a while
#
print('\t\tFlushing extant org data...')
self.sql_driver.reset_domains_orgs()
# next we pull the org/domain pairings from the json file in the resources dir
# and add to the db
print('\t\tPatching with new org data...')
raw_data = open(os.path.join(os.path.dirname(__file__), './resources/org_domains/org_domains.json'), 'r')
json_data = json.load(raw_data)
# the default id for orgs is 1, so we advance from there
id = 1
for item in json_data:
id += 1
self.sql_driver.add_org(id, item['organization'], item['notes'], item['country'])
for domain in item['domains']:
self.sql_driver.update_domain_org(id, domain)
# end patch_org_data
def get_top_tlds(self, limit, type = 'tld'):
# finds the most common tlds from all the pages
# type is default to tld, but pubsuffix also works
# have to do some weird sorting b/c python is arbitrary and f***s up diff tests
# returns array
tlds = []
for row in self.sql_driver.get_all_tlds(type):
tlds.append(row[0])
top_tlds = collections.Counter(tlds).most_common()
# sort alphabetical
top_tlds.sort()
# sub-sort on num occurances
top_tlds.sort(reverse=True, key=lambda item:item[1])
# cut the array
top_tlds = top_tlds[0:limit]
# push in wild cards
top_tlds.insert(0, ('*',self.pages_ok_count))
return top_tlds
# end get_top_tlds
def get_tracker_domains(self, threshold):
# NOTE: first determines all pairings of page domains and element domains
# note this is then unique on SITES, not on PAGES
# e.g. if you have several pages from the same site these links only
# count once
#
# RETURNS: list of domains which link at least the threshold number of sites
all_domains = []
for page_domain_element_domain in self.sql_driver.get_page_domain_element_domain_pairs():
all_domains.append(page_domain_element_domain[1])
# count up all the pairs, convert to items() so can process as tuples
domain_counts = collections.Counter(all_domains).items()
# put the return values here
tracker_domains = []
# check against threshold
for domain_count in domain_counts:
if domain_count[1] >= threshold:
tracker_domains.append(domain_count[0])
# EDGE CASE
# likely due to a large threshold we have no tracker domains,
# so we throw warning and log error
if len(tracker_domains) == 0:
self.sql_driver.log_error('Analaysis Warning', 'Tracker Threshold of %s resulted in no tracking domains.' % threshold)
print('\t\t-----------WARNING-----------')
print('\t\tTracker Threshold of %s resulted in no tracking domains.' % threshold)
print('\t\t-----------------------------')
return tracker_domains
# get_tracker_domains
#########################
# REPORTS, GENERAL #
#########################
def header(self):
# just outputs really basic data about how many records in db, etc.
#
print('\t=================')
print('\t General Summary')
print('\t=================')
output_for_csv = []
#write csv header
output_for_csv.append('"Item","Value"\n')
total_pages_ok = self.sql_driver.pages_ok_count()
print("\t\tTotal Pages OK:\t\t\t%s" % total_pages_ok)
output_for_csv.append('"Total Pages OK","%s"\n' % total_pages_ok)
total_pages_noload = self.sql_driver.pages_noload_count()
total_pages_attempted = total_pages_ok + total_pages_noload
print("\t\tTotal Pages FAIL:\t\t%s" % total_pages_noload)
output_for_csv.append('"Total Pages FAIL","%s"\n' % total_pages_noload)
print("\t\tTotal Pages Attempted:\t\t%s" % total_pages_attempted)
output_for_csv.append('"Total Pages Attempted","%s"\n' % total_pages_attempted)
percent_pages_OK = int((total_pages_ok/total_pages_attempted)*100)
print("\t\t%% Pages OK:\t\t\t%s%%" % percent_pages_OK)
output_for_csv.append('"%% Pages OK","%s"\n' % percent_pages_OK)
total_errors = self.sql_driver.total_errors_count()
print("\t\tTotal Errors:\t\t\t%s" % total_errors)
output_for_csv.append('"Total Errors","%s"\n' % total_errors)
total_cookies = self.sql_driver.total_cookie_count()
print("\t\tTotal Cookies:\t\t\t%s" % total_cookies)
output_for_csv.append('"Total Cookies","%s"\n' % total_cookies)
total_pages_with_cookies = self.sql_driver.pages_w_cookie_count()
print("\t\tPages with Cookies:\t\t%s" % total_pages_with_cookies)
output_for_csv.append('"Pages with Cookies","%s"\n' % total_pages_with_cookies)
percent_with_cookies = (total_pages_with_cookies/total_pages_ok)*100
print("\t\t%% Pages with Cookies:\t\t%s%%" % int(percent_with_cookies))
output_for_csv.append('"%% Pages with Cookies","%s"\n' % int(percent_with_cookies))
total_elements = self.sql_driver.total_element_count()
print("\t\tTotal Elements Requested:\t%s" % total_elements)
output_for_csv.append('"Total Elements Requested","%s"\n' % total_elements)
total_elements_received = self.sql_driver.total_element_count(received = True)
print("\t\tTotal Elements Received:\t%s" % total_elements_received)
output_for_csv.append('"Total Elements Received","%s"\n' % total_elements_received)
percent_element_received = int((total_elements_received/total_elements)*100)
print('\t\t%% Elements Received:\t\t%s%%' % percent_element_received)
output_for_csv.append('"%% Elements Received","%s"\n' % percent_element_received)
total_pages_with_elements = self.sql_driver.pages_w_element_count()
print("\t\tPages with Elements:\t\t%s" % total_pages_with_elements)
output_for_csv.append('"Pages with Elements","%s"\n' % total_pages_with_elements)
percent_with_elements = (total_pages_with_elements/total_pages_ok)*100
print("\t\t%% Pages with Elements:\t\t%s%%" % int(percent_with_elements))
output_for_csv.append('"%% Pages With Elements","%s"\n' % int(percent_with_elements))
self.write_csv('db_summary.csv', output_for_csv)
print('\t'+'-'*80+'\n')
# header
def get_network_ties(self):
# this report generates data necessary for graph/network analysis by
# outputting a list of page domains and the elements/orgs they connect to
print('\t=============================')
print('\t Processing Network Ties ')
print('\t=============================')
# put output here
output_for_csv = []
# header row for csv
output_for_csv.append('"page_domain","3p_org","3p_domain"\n')
# sql_driver.get_network_ties returns a set of tuples in the format
# (page domain, page org, element domain, element org)
# we just go through this data to produce the report
# note: the report is currently omitting page org, but it can be altered easily
for edge in self.sql_driver.get_network_ties():
# if a page has no elements, edge[2] will be 'None' so we skip it
# an alternate approach would be to include as orphan nodes
if edge[2]: output_for_csv.append('"%s","%s","%s",\n' % (edge[0],edge[3],edge[2]))
# done!
self.write_csv('network.csv', output_for_csv)
print('\t'+'-'*80+'\n')
# get_network_ties
def get_summary_by_tld(self):
print('\t=============================')
print('\t Processing Summaries by TLD ')
print('\t=============================')
output_for_csv = []
output_for_csv.append('"TLD","N","% TOTAL","N W/3PE","% W/3PE","N W/COOKIE","% W/COOKIE","N W/JS","% W/JS","3P DOMAIN MEAN","3P DOMAIN MEDIAN","3P DOMAIN MODE","TRACKER FILTER DEPTH"\n')
# now do per-tld numbers
for tld in self.top_tlds:
print('\t\tGetting summary for %s' % tld[0])
if tld[0] != '*':
tld_filter = tld[0]
else:
tld_filter = ''
total_pages = self.sql_driver.get_complex_page_count(tld_filter)
total_pages_percent = (total_pages/self.pages_ok_count)*100
total_pages_elements = self.sql_driver.get_complex_page_count(tld_filter, 'elements', self.tracker_domains)
percent_with_elements = (total_pages_elements/total_pages)*100
total_pages_cookies = self.sql_driver.get_complex_page_count(tld_filter, 'cookies', self.tracker_domains)
percent_with_cookies = (total_pages_cookies/total_pages)*100
total_pages_js = self.sql_driver.get_complex_page_count(tld_filter, 'javascript', self.tracker_domains)
percent_with_js = (total_pages_js/total_pages)*100
if self.tracker_threshold:
filter_depth = self.tracker_threshold
else:
filter_depth = 'No Tracker Filter Used'
stats = self.get_page_3p_stats(tld[0])
mean = stats[0]
median = stats[1]
mode = stats[2]
output_for_csv.append('"%s","%s","%.2f","%s","%.2f","%s","%.2f","%s","%.2f","%.2f","%s","%s","%s"\n' % (
tld[0],
total_pages,
total_pages_percent,
total_pages_elements,
percent_with_elements,
total_pages_cookies,
percent_with_cookies,
total_pages_js,
percent_with_js,
mean,
median,
mode,
filter_depth))
self.write_csv('summary_by_tld.csv', output_for_csv)
# end get_summary_by_tld
#####################
# REPORTS, MAIN #
#####################
def get_reports_by_tld(self, type='', sub_type=''):
print('\t=============================')
print('\tProcessing Top %s %s %s' % (self.num_results, type, sub_type))
print('\t=============================')
# keep output here
csv_output = []
# write out the header row for the csv
if type is 'elements':
csv_output.append('"TLD","TLD Rank","Intra-TLD Rank","Organization","Country", "Element","Extension","Type","Domain","Total Pages","Raw Count","Percent Total"\n')
elif type is 'domains':
csv_output.append('"TLD","TLD Rank","Intra-TLD Rank","Domain","Organization","Country","Number of Pages","Raw Count","Percent Total"\n')
elif type is 'orgs':
csv_output.append('"TLD","TLD Rank","Intra-TLD Rank","Organization","Country","Number of Pages","Raw Count","Percent Total"\n')
else:
self.fatal('Wrong type specified in get_reports_by_tld, must be "elements", "domains", or "orgs".')
tld_count = 0
for tld in self.top_tlds:
current_tld = tld[0]
total_pages = tld[1]
print('\t\tcurrently on: '+current_tld)
# filter on current page tld
if tld[0] != '*':
tld_filter = tld[0]
tld_count += 1
else:
tld_filter = ''
# get results with specified filter
results_rows = self.get_results_rows(total_pages, type, sub_type, tld_filter)
current_row = 0
# loop results
for result_row in results_rows:
current_row += 1
if type is 'elements':
csv_row = '"%s","%s","%s","%s","%s","%s","%s","%s","%s","%s","%s","%.2f"\n' % (
current_tld,
tld_count,
current_row,
result_row[0][0], # org_name
result_row[0][1], # country
result_row[0][2], # element
result_row[0][3], # extension
result_row[0][4], # type
result_row[0][5], # domain
total_pages,
result_row[1], # raw_count
(result_row[1]/total_pages)*100)
elif type is 'domains':
total_item = result_row[1]
csv_row = '"%s","%s","%s","%s","%s","%s","%s","%s","%.2f"\n' % (
current_tld,
tld_count,
current_row,
result_row[0][0], # domain_name
result_row[0][1], # org_name
result_row[0][2], # org_country
total_pages,
total_item,
(total_item/total_pages)*100)
elif type is 'orgs':
total_item = result_row[1]
csv_row = '"%s","%s","%s","%s","%s","%s","%s","%.2f"\n' % (
current_tld,
tld_count,
current_row,
result_row[0][0], # org_name
result_row[0][1], # org_country
total_pages,
total_item,
(total_item/total_pages)*100)
csv_output.append(csv_row)
if current_row >= self.num_results: break
# write out csv
file_name = type + '-by-tld'
# this really only applied to elements at present
if sub_type:
file_name += '-' + sub_type
file_name += '.csv'
# store the report
self.write_csv(file_name, csv_output)
# end get_reports_by_tld
def get_results_rows(self, total_pages, type, sub_type = '', tld_filter=''):
# this queries the db to get all elements, domains, or orgs
# next they are counted to find the most common
# and formatted to csv rows and returned
# query the db
if type is 'elements':
# rows are (page.start_uri, org.name, element.element_uri, element.extension, element.type, element_domain.domain)
query_results = self.sql_driver.get_elements(tld_filter, sub_type)
elif type is 'domains':
# rows are (page.start_uri, element_domain.domain, org.name)
query_results = self.sql_driver.get_domains(tld_filter)
elif type is 'orgs':
# row are page.start_uri, org.name
query_results = self.sql_driver.get_orgs(tld_filter)
else:
self.fatal('Type must be elements, domains, or orgs.')
# count up the unique elements, domains, or orgs we are looking for
results_counter = collections.Counter()
for row in query_results:
# remove first element in tuple as it is the page uri, now irrelevant
# add rest to results counter as this is what we care about now
results_counter[row[1:]] += 1
# python's most_common() arbitrarily re-orders items with same value, making
# debugging a nightmare, so we have to double sort here
# convert to list we can sort
results_counter = results_counter.most_common()
# sort alphabetical
results_counter.sort()
# sub-sort on num occurrences
results_counter.sort(reverse=True, key=lambda item:item[1])
return results_counter
# end get_results_rows
def get_page_3p_stats(self, tld = ''):
# This function calls get_page_element_domain_pairs to get a list of tuples
# st. each tuple is a unique (page address, domain of an element) paring
# This list of tuples is then iterated so that we count how many domains
# each page is linked to
# IMPORTANT: get_page_element_domain_pairs is *already* sorted by page, otherwise
# loop below would not work
if tld == '*':
tld = ''
# init vars
this_page_element_count = 0
all_page_element_counts = []
last_page = ''
# run query, process rows
for row in self.sql_driver.get_page_uri_element_domain_pairs(tld):
# page has no tackers, count is zero
if not row[1]:
all_page_element_counts.append(0)
# page has trackers, add count
else:
# this is the same page, increment count
if row[0] == last_page:
if self.tracker_domains:
if row[1] in self.tracker_domains:
this_page_element_count += 1
else:
this_page_element_count += 1
# this is a new page, store our running count, reset to 1
# update last_page
else:
if last_page != '': all_page_element_counts.append(this_page_element_count)
last_page = row[0]
if self.tracker_domains:
if row[1] in self.tracker_domains:
this_page_element_count = 1
else:
this_page_element_count = 0
else:
this_page_element_count = 1
# if we have an outstanding value, give it an increment
if this_page_element_count != 0:
# enter the last value into the list
all_page_element_counts.append(this_page_element_count)
# mean and median should always be ok
mean = statistics.mean(all_page_element_counts)
median = statistics.median(all_page_element_counts)
# but mode can throw an error, so catch here
try:
mode = statistics.mode(all_page_element_counts)
except:
mode = 'NULL'
return(mean, median, mode)
# get_page_3p_stats
# end class Reporter
class ParseURI:
def __init__(self):
# load up the tld list now as only hit it once this way
self.pubsuffix_list = self.get_pubsuffix_list()
# this is to speed up tlds lookup with hash table
# we can share among all runs over time
self.sql_driver = MySQLDriver()
# check if sub_domain_tld exists, if not create one
# this should only really happen once
#
# bug?
# when using pool this can be done several times in tandem
# resulting in some lost entries - e.g. two threads see db not exist the first
# one to break the tie will create a new blank db and add a new entry
# it is possible the slightly slower second thread will also create a blank db
# erasing the entry of the first thread, however, after the first round the db
# should not be re-created again
#
# it should be emphasized this db is ONLY used to speed up parsing the tld
# (partially because my regex method is slow and should be refactored!)
# the data is this db is NOT used for analysis - thus deleting a few records
# will have a minor impact on speed for the first few entries, and then be
# wholly irrelevant
#
# still...it irks me that this entire class could be smarter, so I will
# fix it another day
if self.sql_driver.check_db_exist('sub_domain_tld') == False:
self.sql_driver.create_sub_domain_tld_db()
else:
self.sql_driver.db_switch('sub_domain_tld')
# end __init__
def get_pubsuffix_list(self):
# get the file from the local dir
pubsuffix_raw_list = open(os.path.join(os.path.dirname(__file__), './resources/pubsuffix/wbxrPubSuffixList.txt'), 'r')
pubsuffix_list = []
for line in pubsuffix_raw_list:
# the last part of the list is random shit we don't care about, so stop reading
if re.match("^// ===BEGIN PRIVATE DOMAINS===", line):break
# skip lines that are comments or blank, add others to list
# also remove leading ., !, and * as it f***s up regex later
if not re.match("^//.+$|^$", line):
pubsuffix_list.append(re.sub('^[\!\*]\.?', '', line.strip()))
# we sort long->short so we can match deeper TLD first (ie, ac.uk *before* .uk)
pubsuffix_list.sort(key=len,reverse=True)
# to speed things up we move the most common TLD to the front of the line
# that said, if it's not one of these we take a MASSIVE performance hit
popular_pubsuffixs = ['gov', 'co.uk', 'mil', 'org', 'net', 'edu', 'com']
for popular_pubsuffix in popular_pubsuffixs:
pubsuffix_list.remove(popular_pubsuffix)
pubsuffix_list.insert(0, popular_pubsuffix)
return pubsuffix_list
# get_pubsuffix_list
def get_domain_pubsuffix_tld(self, uri):
# pull out the first chunk of the domain, possibly including subdomains
# if this fails, the domain is f****d up (or not https? at least), kill
# only handles if the domain is followed by $, ?, \, or / - other shit will break...
# adding to handle with port at end, ie blah.com:8080, so rule is '|\:[0-9].+
# adding \.? at the top to handle leading '.'
try:
sub_domain = re.search('^https?:\/\/\.?(.*?)(\:[0-9].+)?($|\/|\?|\\\\|=)', uri).group(1)
except:
return('Exception: Unable to parse: '+uri[:50], 'Exception: Unable to parse: '+uri[:50], 'Exception: Unable to parse: '+uri[:50])
# strip off leading or trailing '.', this is f*****g things up
sub_domain = re.sub('(^\.|\.$)', '', sub_domain)
# see if it is an ip address, if so return it
# this is only ipv4 pattern though, maybe should thinking about ipv6?
try:
re.search('(^\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}$)', sub_domain).group(1)
return (sub_domain, 'None', 'None')
except:
pass
# first, we will see if it is in the db already
record = self.sql_driver.sub_domain_exists(sub_domain)
if record:
return record
# the pubsuffix list is large -> small,
# so the first match is the one we want
# after we find it, break out and continue
for pubsuffix_try in self.pubsuffix_list:
if re.match(".+\."+pubsuffix_try+"$", sub_domain):
pubsuffix = pubsuffix_try
break
# if we didn't find the pubsuffix we fail
try:
pubsuffix
except:
return('Exception: Unable to parse: '+uri[:50], 'Exception: Unable to parse: '+uri[:50], 'Exception: Unable to parse: '+uri[:50])
# if we have sub.domain.tld_match, we just want domain.tld_match
# there is no reason this should fail if we get this far, but try/except to be safe
try:
domain = re.search('(.*\.)?\.?(.*\.'+pubsuffix+')$', sub_domain).group(2)
except:
return('Exception: Unable to parse: '+uri[:50], 'Exception: Unable to parse: '+uri[:50], 'Exception: Unable to parse: '+uri[:50])
# grab the tld off of the pubsuffix
# if regex fails the tld and pubsuffix are the same
try:
tld = re.search('\.([0-9A-Za-z]+)$', pubsuffix).group(1).lower()
except:
tld = pubsuffix
self.sql_driver.add_sub_domain_pubsuffix_tld(sub_domain, domain, pubsuffix, tld)
return (domain, pubsuffix, tld)
#end get_domain_pubsuffix_tld
def close(self):
# close mysql connection
self.sql_driver.close()
return
# end close
#end ParseURI
class Reporter:
def __init__(self, db_name, num_tlds, num_results, tracker_threshold=0):
self.db_name = db_name
self.sql_driver = MySQLDriver(self.db_name)
self.num_tlds = num_tlds
self.num_results = num_results
self.tracker_threshold = tracker_threshold
self.startTime = datetime.now()
self.pages_ok_count = self.sql_driver.pages_ok_count()
print('\t=============================')
print('\t Checking Output Directories ')
print('\t=============================')
self.setup_report_dir()
print('\t===========================')
print('\t Patching DB with Org Data ')
print('\t===========================')
# update the domains to their owners
self.patch_org_data()
print('\t\tSuccess!')
print('\t=====================')
print('\t Getting top %s tlds' % self.num_tlds)
print('\t=====================')
print('\t\tProcessing...')
self.top_tlds = self.get_top_tlds(self.num_tlds)
print(self.top_tlds)
print('\t\tSuccess!')
print('\t\tThe top tlds are:')
for (tld, pages) in self.top_tlds:
print('\t\t |- %s (%s)' % (tld, pages))
# SPECIAL SAUCE, FOR EXPERTS: tracker domains!
#
# idea for this is you set a threshold of the number of sites a given domain
# is connected to - domains connecting to many sites may correlate those visits
# via referer strings etc, so we call these 'tracker domains'
#
# on a really large set of sites (e.g. 1M+) this works well but on small samples
# (e.g. 500) it doesn't work well at all as known tracker domains may only
# appear on a single site
#
# this is off by default and unless you understand what you are doing...
# DON'T USE THIS!
#
# longer-term we may want to train off a bigger corpus to find tracker domains and
# have them prepackaged
#
if tracker_threshold:
print('\t=========================')
print('\t Getting tracker domains ')
print('\t=========================')
print('\t\tProcessing...')
self.tracker_domains = self.get_tracker_domains(
self.tracker_threshold)
print('\t\tSuccess!')
else:
self.tracker_domains = []
# end __init__
#########################
# HELPERS, GENERAL #
#########################
def setup_report_dir(self):
# create directory for where the reports go if not exist
if os.path.exists('./reports') == False:
print('\tMaking reporting directory.')
os.makedirs('./reports')
# set global report_path, trim off the wbxr_ prefix
self.report_path = './reports/' + self.db_name[5:]
# set up subdir for this database analysis
if os.path.exists(self.report_path) == False:
print('\tMaking subdirectory for reports.')
os.makedirs(self.report_path)
# just a notice
print('\t\tWriting output to %s' % self.report_path)
# setup_report_dir
# reuse this a lot
def write_csv(self, file_name, csv_rows):
full_file_path = self.report_path + '/' + file_name
file_out = open(full_file_path, 'w')
for row in csv_rows:
file_out.write(row)
file_out.close()
print('\t\t' + '*' * 40)
print('\t\tOutput written to %s' % full_file_path)
# write_csv
# just fyi
def print_runtime(self):
print('~=' * 40)
print("Finished!")
print("Time to process: " + str(datetime.now() - self.startTime) +
"\n")
print('-' * 80)
# end print_runtime
# X-(
def fatal(self, msg):
print('FATAL ERROR: %s' % msg)
print('EXITING.')
exit()
# fatal
#############################
# HELPERS, DATABASE/INIT #
#############################
def patch_org_data(self):
# in order to analyze what entities receive user data, we need to update
# the database with domain ownership records we have store previously
#
# we first clear out what is in there in case the new data has changed
# perhaps make this optional, on big dbs takes a while
#
print('\t\tFlushing extant org data...')
self.sql_driver.reset_domains_orgs()
# next we pull the org/domain pairings from the json file in the resources dir
# and add to the db
print('\t\tPatching with new org data...')
raw_data = open(
os.path.join(os.path.dirname(__file__),
'./resources/org_domains/org_domains.json'), 'r')
json_data = json.load(raw_data)
# the default id for orgs is 1, so we advance from there
id = 1
for item in json_data:
id += 1
self.sql_driver.add_org(id, item['organization'], item['notes'],
item['country'])
for domain in item['domains']:
self.sql_driver.update_domain_org(id, domain)
# end patch_org_data
def get_top_tlds(self, limit, type='tld'):
# finds the most common tlds from all the pages
# type is default to tld, but pubsuffix also works
# have to do some weird sorting b/c python is arbitrary and f***s up diff tests
# returns array
tlds = []
for row in self.sql_driver.get_all_tlds(type):
tlds.append(row[0])
top_tlds = collections.Counter(tlds).most_common()
# sort alphabetical
top_tlds.sort()
# sub-sort on num occurances
top_tlds.sort(reverse=True, key=lambda item: item[1])
# cut the array
top_tlds = top_tlds[0:limit]
# push in wild cards
top_tlds.insert(0, ('*', self.pages_ok_count))
return top_tlds
# end get_top_tlds
def get_tracker_domains(self, threshold=0):
# first finds all pairings of page domains and element domains
# note this is then unique on SITE, not on PAGE
# returns the list of domains which appear on at least the threshold number
domains = []
for page_domain_element_domain in self.sql_driver.get_page_domain_element_domain_pairs(
):
domains.append(page_domain_element_domain[1])
# count up all the pairs, convert to items() so can process as tuples
domain_counts = collections.Counter(domains).items()
# put the return values here
tracker_domains = []
# check against threshold
for domain_count in domain_counts:
if domain_count[1] >= threshold:
tracker_domains.append(domain_count[0])
return tracker_domains
# get_tracker_domains
#########################
# REPORTS, GENERAL #
#########################
def header(self):
# just outputs really basic data about how many records in db, etc.
#
print('\t=================')
print('\t General Summary')
print('\t=================')
output_for_csv = []
#write csv header
output_for_csv.append('"Item","Value"\n')
total_pages = self.sql_driver.pages_ok_count()
print("\t\tTotal Pages OK:\t\t\t%s" % total_pages)
output_for_csv.append('"Total Pages OK","%s"\n' % total_pages)
total_pages_noload = self.sql_driver.pages_noload_count()
total_pages_attempted = total_pages + total_pages_noload
print("\t\tTotal Pages FAIL:\t\t%s" % total_pages_noload)
output_for_csv.append('"Total Pages FAIL","%s"\n' % total_pages_noload)
print("\t\tTotal Pages Attempted:\t\t%s" % total_pages_attempted)
output_for_csv.append('"Total Pages Attempted","%s"\n' %
total_pages_attempted)
percent_pages_OK = int((total_pages / total_pages_attempted) * 100)
print("\t\t%% Pages OK:\t\t\t%s%%" % percent_pages_OK)
output_for_csv.append('"%% Pages OK","%s"\n' % percent_pages_OK)
total_errors = self.sql_driver.total_errors_count()
print("\t\tTotal Errors:\t\t\t%s" % total_errors)
output_for_csv.append('"Total Errors","%s"\n' % total_errors)
total_cookies = self.sql_driver.total_cookie_count()
print("\t\tTotal Cookies:\t\t\t%s" % total_cookies)
output_for_csv.append('"Total Cookies","%s"\n' % total_cookies)
total_pages_with_cookies = self.sql_driver.pages_w_cookie_count()
print("\t\tPages with Cookies:\t\t%s" % total_pages_with_cookies)
output_for_csv.append('"Pages with Cookies","%s"\n' %
total_pages_with_cookies)
percent_with_cookies = (total_pages_with_cookies / total_pages) * 100
print("\t\t%% Pages with Cookies:\t\t%s%%" % int(percent_with_cookies))
output_for_csv.append('"%% Pages with Cookies","%s"\n' %
int(percent_with_cookies))
total_elements = self.sql_driver.total_element_count()
print("\t\tTotal Elements Requested:\t%s" % total_elements)
output_for_csv.append('"Total Elements Requested","%s"\n' %
total_elements)
total_elements_received = self.sql_driver.total_element_count(
received=True)
print("\t\tTotal Elements Received:\t%s" % total_elements_received)
output_for_csv.append('"Total Elements Received","%s"\n' %
total_elements_received)
percent_element_received = int(
(total_elements_received / total_elements) * 100)
print('\t\t%% Elements Received:\t\t%s%%' % percent_element_received)
output_for_csv.append('"%% Elements Received", "%s"\n' %
percent_element_received)
total_pages_with_elements = self.sql_driver.pages_w_element_count()
print("\t\tPages with Elements:\t\t%s" % total_pages_with_elements)
output_for_csv.append('"Pages with Elements","%s"\n' %
total_pages_with_elements)
percent_with_elements = (total_pages_with_elements / total_pages) * 100
print("\t\t%% Pages with Elements:\t\t%s%%" %
int(percent_with_elements))
output_for_csv.append('"%% Pages With Elements","%s"\n' %
int(percent_with_elements))
self.write_csv('db_summary.csv', output_for_csv)
print('\t' + '-' * 80 + '\n')
# header
def get_network_ties(self):
print('\t=============================')
print('\t Processing Network Ties ')
print('\t=============================')
output_for_csv = []
# can also include the page_org in the report, but commented out for now
# at a later date this could be an option
# output_for_csv.append('"page_domain","page_org","3p_domain","3p_org"\n')
output_for_csv.append('"page_domain","3p_domain","3p_org"\n')
for edge in self.sql_driver.get_network_ties():
# output_for_csv.append('"%s","%s","%s","%s"\n' % (edge[0],edge[1],edge[2],edge[3]))
output_for_csv.append('"%s","%s","%s",\n' %
(edge[0], edge[2], edge[3]))
self.write_csv('network.csv', output_for_csv)
print('\t' + '-' * 80 + '\n')
# get_network_ties
def get_summary_by_tld(self):
print('\t=============================')
print('\t Processing Summaries by TLD ')
print('\t=============================')
output_for_csv = []
output_for_csv.append(
'"TLD","N","% TOTAL","N W/3PE","% W/3PE","N W/COOKIE","% W/COOKIE","N W/JS","% W/JS","3P DOMAIN MEAN","3P DOMAIN MEDIAN","3P DOMAIN MODE" \n'
)
# now do per-tld numbers
for tld in self.top_tlds:
print('\t\tGetting summary for %s' % tld[0])
if tld[0] != '*':
tld_filter = tld[0]
else:
tld_filter = ''
total_pages = self.sql_driver.get_complex_page_count(tld_filter)
total_pages_percent = (total_pages / self.pages_ok_count) * 100
total_pages_elements = self.sql_driver.get_complex_page_count(
tld_filter, 'elements', self.tracker_domains)
percent_with_elements = (total_pages_elements / total_pages) * 100
total_pages_cookies = self.sql_driver.get_complex_page_count(
tld_filter, 'cookies', self.tracker_domains)
percent_with_cookies = (total_pages_cookies / total_pages) * 100
total_pages_js = self.sql_driver.get_complex_page_count(
tld_filter, 'javascript', self.tracker_domains)
percent_with_js = (total_pages_js / total_pages) * 100
stats = self.get_page_3p_stats(tld[0])
mean = stats[0]
median = stats[1]
mode = stats[2]
output_for_csv.append(
'"%s","%s","%.2f","%s","%.2f","%s","%.2f","%s","%.2f","%.2f","%s","%s"\n'
% (tld[0], total_pages, total_pages_percent,
total_pages_elements, percent_with_elements,
total_pages_cookies, percent_with_cookies, total_pages_js,
percent_with_js, mean, median, mode))
self.write_csv('summary_by_tld.csv', output_for_csv)
# end get_summary_by_tld
#####################
# REPORTS, MAIN #
#####################
def get_reports_by_tld(self, type='', sub_type=''):
print('\t=============================')
print('\tProcessing Top %s %s %s' % (self.num_results, type, sub_type))
print('\t=============================')
# keep output here
csv_output = []
# write out the header row for the csv
if type is 'elements':
csv_output.append(
'"TLD","TLD Rank","Intra-TLD Rank","Organization","Country", "Element","Extension","Type","Domain","Total Pages","Raw Count","Percent Total"\n'
)
elif type is 'domains':
csv_output.append(
'"TLD","TLD Rank","Intra-TLD Rank","Domain","Organization","Country","Number of Pages","Raw Count","Percent Total"\n'
)
elif type is 'orgs':
csv_output.append(
'"TLD","TLD Rank","Intra-TLD Rank","Organization","Country","Number of Pages","Raw Count","Percent Total"\n'
)
else:
self.fatal(
'Wrong type specified in get_reports_by_tld, must be "elements", "domains", or "orgs".'
)
tld_count = 0
for tld in self.top_tlds:
current_tld = tld[0]
total_pages = tld[1]
print('\t\tcurrently on: ' + current_tld)
# filter on current page tld
if tld[0] != '*':
tld_filter = tld[0]
tld_count += 1
else:
tld_filter = ''
# get results with specified filter
results_rows = self.get_results_rows(total_pages, type, sub_type,
tld_filter)
current_row = 0
# loop results
for result_row in results_rows:
current_row += 1
if type is 'elements':
csv_row = '"%s","%s","%s","%s","%s","%s","%s","%s","%s","%s","%s","%.2f"\n' % (
current_tld,
tld_count,
current_row,
result_row[0][0], # org_name
result_row[0][1], # country
result_row[0][2], # element
result_row[0][3], # extension
result_row[0][4], # type
result_row[0][5], # domain
total_pages,
result_row[1], # raw_count
(result_row[1] / total_pages) * 100)
elif type is 'domains':
total_item = result_row[1]
csv_row = '"%s","%s","%s","%s","%s","%s","%s","%s","%.2f"\n' % (
current_tld,
tld_count,
current_row,
result_row[0][0], # domain_name
result_row[0][1], # org_name
result_row[0][2], # org_country
total_pages,
total_item,
(total_item / total_pages) * 100)
elif type is 'orgs':
total_item = result_row[1]
csv_row = '"%s","%s","%s","%s","%s","%s","%s","%.2f"\n' % (
current_tld,
tld_count,
current_row,
result_row[0][0], # org_name
result_row[0][1], # org_country
total_pages,
total_item,
(total_item / total_pages) * 100)
csv_output.append(csv_row)
if current_row >= self.num_results: break
# write out csv
file_name = type + '-by-tld'
# this really only applied to elements at present
if sub_type:
file_name += '-' + sub_type
file_name += '.csv'
# store the report
self.write_csv(file_name, csv_output)
# end get_reports_by_tld
def get_results_rows(self, total_pages, type, sub_type='', tld_filter=''):
# this queries the db to get all elements, domains, or orgs
# next they are counted to find the most common
# and formatted to csv rows and returned
# query the db
if type is 'elements':
# rows are (page.start_uri, org.name, element.element_uri, element.extension, element.type, element_domain.domain)
query_results = self.sql_driver.get_elements(tld_filter, sub_type)
elif type is 'domains':
# rows are (page.start_uri, element_domain.domain, org.name)
query_results = self.sql_driver.get_domains(tld_filter)
elif type is 'orgs':
# row are page.start_uri, org.name
query_results = self.sql_driver.get_orgs(tld_filter)
else:
self.fatal('Type must be elements, domains, or orgs.')
# count up the unique elements, domains, or orgs we are looking for
results_counter = collections.Counter()
for row in query_results:
# remove first element in tuple as it is the page uri, now irrelevant
# add rest to results counter as this is what we care about now
results_counter[row[1:]] += 1
# python's most_common() arbitrarily re-orders items with same value, making
# debugging a nightmare, so we have to double sort here
# convert to list we can sort
results_counter = results_counter.most_common()
# sort alphabetical
results_counter.sort()
# sub-sort on num occurrences
results_counter.sort(reverse=True, key=lambda item: item[1])
return results_counter
# end get_results_rows
def get_page_3p_stats(self, tld=''):
# This function calls get_page_element_domain_pairs to get a list of tuples
# st. each tuple is a unique (page address, domain of an element) paring
# This list of tuples is then iterated so that we count how many domains
# each page is linked to
# IMPORTANT: get_page_element_domain_pairs is *already* sorted by page, otherwise
# loop below would not work
if tld == '*':
tld = ''
# init vars
this_page_element_count = 0
all_page_element_counts = []
last_page = ''
# run query, process rows
for row in self.sql_driver.get_page_uri_element_domain_pairs(tld):
# page has no tackers, count is zero
if not row[1]:
all_page_element_counts.append(0)
# page has trackers, add count
else:
# this is the same page, increment count
if row[0] == last_page:
if self.tracker_domains:
if row[1] in self.tracker_domains:
this_page_element_count += 1
else:
this_page_element_count += 1
# this is a new page, store our running count, reset to 1
# update last_page
else:
if last_page != '':
all_page_element_counts.append(this_page_element_count)
last_page = row[0]
if self.tracker_domains:
if row[1] in self.tracker_domains:
this_page_element_count = 1
else:
this_page_element_count = 0
else:
this_page_element_count = 1
# if we have an outstanding value, give it an increment
if this_page_element_count != 0:
# enter the last value into the list
all_page_element_counts.append(this_page_element_count)
# mean and median should always be ok
mean = statistics.mean(all_page_element_counts)
median = statistics.median(all_page_element_counts)
# but mode can throw an error, so catch here
try:
mode = statistics.mode(all_page_element_counts)
except:
mode = 'NULL'
return (mean, median, mode)
# get_page_3p_stats
# end class Reporter
def process_url(self, url):
"""
this function takes a specified url, loads it in the browser (currently phantomjs)
and returns json-formatted output with relevant request data, etc.
the output_store class then puts this data in the db for later analysis
"""
# set up sql connection used to log errors and do timeseries checks
if self.db_engine == 'mysql':
from webxray.MySQLDriver import MySQLDriver
sql_driver = MySQLDriver(self.db_name)
elif self.db_engine == 'postgres':
from webxray.PostgreSQLDriver import PostgreSQLDriver
sql_driver = PostgreSQLDriver(self.db_name)
elif self.db_engine == 'sqlite':
from webxray.SQLiteDriver import SQLiteDriver
sql_driver = SQLiteDriver(self.db_name)
# output store does the heavy lifting of analyzing browser output and storing to db
output_store = OutputStore(self.db_engine, self.db_name)
# support for loading same page with multiple browsers - purposefully undocumented
for browser_type in self.browser_types:
# import and set up specified browser driver
# note we need to set up a new browser each time to
# get a fresh profile
if browser_type == 'phantomjs':
browser_driver = PhantomDriver()
elif browser_type == 'chrome':
browser_driver = ChromeDriver(ua=self.chrome_ua)
# support for timeseries collections - purposefully undocumented
if self.allow_timeseries:
page_last_accessed_browser_type = sql_driver.get_page_last_accessed_by_browser_type(url,browser_type)
if page_last_accessed_browser_type:
time_diff = datetime.now()-page_last_accessed_browser_type[0]
if time_diff < timedelta(minutes=self.interval_minutes) and page_last_accessed_browser_type[1] == browser_type:
print("\t\t%-50s Scanned too recently with %s" % (url[:50], browser_type))
continue
# attempt to load the page, fail gracefully
try:
browser_output = browser_driver.get_webxray_scan_data(url, self.browser_wait)
except:
print('\t\t%-50s Browser %s Did Not Return' % (url[:50], browser_type))
sql_driver.log_error(url, 'Unable to load page')
sql_driver.close()
return
# if there was a problem we log the error
if browser_output['success'] == False:
print('\t\t%-50s Browser %s Error: %s' % (url[:50], browser_type, browser_output['result']))
sql_driver.log_error(url, 'Unable to load page')
sql_driver.close()
return
else:
# no error, treat result as browser output
browser_output = browser_output['result']
# attempt to store the output
if output_store.store(url, browser_output):
print('\t\t%-50s Success with %s' % (url[:50],browser_type))
else:
print('\t\t%-50s Fail with %s' % (url[:50],browser_type))
sql_driver.log_error(url, 'Unable to load page')
sql_driver.close()
return
def run(self, pool_size):
try:
uri_list = open('./page_lists/'+self.pages_file_name, 'r')
except:
print('File "%s" does not exist, file must be in ./page_lists directory. Exiting.' % self.pages_file_name)
exit()
sql_driver = MySQLDriver(self.db_name)
# sort out what uris we are processing from the list
uris_to_process = []
count = 0
print('\t------------------------')
print('\t Building List of Pages ')
print('\t------------------------')
for uri in uri_list:
# skip lines that are comments
if "#" in uri[0]: continue
count += 1
# only do lines starting with https?://
if not (re.match('^https?://.+', uri)):
print("\t\t%s | %-50s Not a valid address, Skipping." % (count, uri[:50]))
continue
# non-ascii domains will crash phantomjs, so we need to convert them to
# idna/ascii/utf-8
# this requires splitting apart the uri, converting the domain to idna,
# and pasting it all back together. ugly.
parsed_uri = urlsplit(uri.strip())
uri = parsed_uri[0] + "://"
uri += parsed_uri[1].encode('idna').decode('utf-8')
# if chunks exist glue them back together
if len(parsed_uri[2]) != 0:
uri += parsed_uri[2]
if len(parsed_uri[3]) != 0:
uri += '?' + parsed_uri[3]
if len(parsed_uri[4]) != 0:
uri += '#' + parsed_uri[4]
# if it is a m$ office or other doc, skip
if re.match('.+(pdf|ppt|pptx|doc|docx|txt|rtf|xls|xlsx)$', uri):
print("\t\t%s | %-50s Not an HTML document, Skipping." % (count, uri[:50]))
continue
# skip if in db already
if sql_driver.page_exists(uri):
print("\t\t%s | %-50s Exists in DB, Skipping." % (count, uri[:50]))
continue
# only add if not in list already
if uri not in uris_to_process:
print("\t\t%s | %-50s Adding." % (count, uri[:50]))
uris_to_process.append(uri)
else:
print("\t\t%s | %-50s Already queued, Skipping." % (count, uri[:50]))
print('\t----------------------------------')
print('\t%s addresses will now be webXray\'d' % len(uris_to_process))
print('\t\t...you can go take a walk. ;-)')
print('\t----------------------------------')
myPool = Pool(pool_size)
myPool.map(self.process_uri, uris_to_process)
def store(self,
url,
browser_output,
store_source=False,
store_1p=True,
get_file_hashes=False,
hash_3p_only=False):
"""
this is the primary function of this class,
it takes the url of the given page and the request and cookie data generated
by the browser
data is cleaned up with some minor analysis (eg file types) and stored
for later in-depth analysis.
there is an option to store first party requests as well as third, turned on by default
to save disk space turn off store_1p
there is also an option to get file hashes, this introduces serious overhead
and is turned off by default
"""
# open up a sql connection
if self.db_engine == 'mysql':
from webxray.MySQLDriver import MySQLDriver
sql_driver = MySQLDriver(self.db_name)
elif self.db_engine == 'sqlite':
from webxray.SQLiteDriver import SQLiteDriver
sql_driver = SQLiteDriver(self.db_name)
elif self.db_engine == 'postgres':
from webxray.PostgreSQLDriver import PostgreSQLDriver
sql_driver = PostgreSQLDriver(self.db_name)
else:
print('INVALED DB ENGINE FOR %s, QUITTING!' % db_engine)
exit()
# get the ip, fqdn, domain, pubsuffix, and tld
# we need the domain to figure out if cookies/elements are third-party
origin_ip_fqdn_domain_pubsuffix_tld = self.url_parser.get_ip_fqdn_domain_pubsuffix_tld(
url)
# if we can't get page domain info we fail gracefully
if origin_ip_fqdn_domain_pubsuffix_tld is None:
sql_driver.log_error(url, 'Could not parse TLD for %s' % url)
return False
origin_ip = origin_ip_fqdn_domain_pubsuffix_tld[0]
origin_fqdn = origin_ip_fqdn_domain_pubsuffix_tld[1]
origin_domain = origin_ip_fqdn_domain_pubsuffix_tld[2]
origin_pubsuffix = origin_ip_fqdn_domain_pubsuffix_tld[3]
origin_tld = origin_ip_fqdn_domain_pubsuffix_tld[4]
# sql_driver.add_domain both stores the new domain and returns its db row id
# if it is already in db just return the existing id
page_domain_id = sql_driver.add_domain(origin_ip, origin_fqdn,
origin_domain, origin_pubsuffix,
origin_tld)
# figure out the privacy policy url and text, starts null
priv_policy_url = None
priv_policy_url_text = None
# read in our list of privacy link terms from the json file in webxray/resources/policyxray
privacy_policy_term_list = self.utilities.get_privacy_policy_term_list(
)
# we reverse links return from browser to check footer links first as that is where policy links tend to be
all_links = browser_output['all_links']
all_links.reverse()
# if we have links search for privacy policy
if len(all_links) > 0:
# links are tuple
for link_text, link_url in all_links:
# makes sure we have text, skip links without
if link_text:
# need lower for string matching
link_text = link_text.lower().strip()
# not a link we can use
if 'javascript' in link_text: continue
# see if the link_text is in our term list
if link_text in privacy_policy_term_list:
# if the link_url is relative this will convert to absolute
priv_policy_url = self.utilities.get_absolute_url_from_page_link(
url, link_url)
priv_policy_url_text = link_text
break
# if the final page is https (often after a redirect), mark it appropriately
if browser_output['final_url'][:5] == 'https':
page_is_ssl = True
else:
page_is_ssl = False
if store_source:
# handles issue where postgres will crash on inserting null character
source = browser_output['source'].replace('\x00', ' ')
else:
source = None
# add page
page_id = sql_driver.add_page(
browser_output['browser_type'], browser_output['browser_version'],
browser_output['browser_wait'], browser_output['title'],
browser_output['meta_desc'], url, browser_output['final_url'],
priv_policy_url, priv_policy_url_text, page_is_ssl, source,
browser_output['load_time'], page_domain_id)
# store cookies
for cookie in browser_output['cookies']:
# get the ip, fqdn, domain, pubsuffix, and tld
# we need the domain to figure out if cookies/elements are third-party
# note:
# url_parser fails on non-http, we should fix this, right now a lame hack is to prepend http://
cookie_ip_fqdn_domain_pubsuffix_tld = self.url_parser.get_ip_fqdn_domain_pubsuffix_tld(
'http://' + cookie['domain'])
# something went wrong, log and fail gracefully
if cookie_ip_fqdn_domain_pubsuffix_tld is None:
sql_driver.log_error(
url,
'Error parsing cookie with domain: ' + cookie['domain'])
continue
# otherwise, everything went fine
cookie_ip = cookie_ip_fqdn_domain_pubsuffix_tld[0]
cookie_fqdn = cookie_ip_fqdn_domain_pubsuffix_tld[1]
cookie_domain = cookie_ip_fqdn_domain_pubsuffix_tld[2]
cookie_pubsuffix = cookie_ip_fqdn_domain_pubsuffix_tld[3]
cookie_tld = cookie_ip_fqdn_domain_pubsuffix_tld[4]
# mark third-party cookies
if origin_domain != cookie_domain:
is_3p_cookie = True
else:
is_3p_cookie = False
# this is a first party cookie, see if we want to store it
if is_3p_cookie is False and store_1p is False:
continue
# sql_driver.add_domain both stores the new domain and returns its id
cookie_domain_id = sql_driver.add_domain(cookie_ip, cookie_fqdn,
cookie_domain,
cookie_pubsuffix,
cookie_tld)
# name and domain are required, so if they fail we just continue
try:
name = cookie['name']
except:
continue
try:
domain = cookie_domain
except:
continue
# these are optional, fill with null values if fail
try:
secure = cookie['secure']
except:
secure = None
try:
path = cookie['path']
except:
path = None
try:
httponly = cookie['httponly']
except:
httponly = None
try:
expiry = cookie['expiry']
except:
expiry = None
try:
value = cookie['value']
except:
value = None
# all done with this cookie
sql_driver.add_cookie(page_id, name, secure, path, domain,
httponly, expiry, value, is_3p_cookie,
cookie_domain_id)
# process requests now
for request in browser_output['processed_requests']:
# if the request starts with the following we can't parse anyway, so skip
if re.match('^(data|about|chrome|blob).+', request):
continue
# get the ip, fqdn, domain, pubsuffix, and tld
# we need the domain to figure out if cookies/elements are third-party
element_ip_fqdn_domain_pubsuffix_tld = self.url_parser.get_ip_fqdn_domain_pubsuffix_tld(
request)
# problem with this request, log and fail gracefully
if element_ip_fqdn_domain_pubsuffix_tld is None:
sql_driver.log_error(
url, 'Error parsing element request: ' + request)
continue
element_ip = element_ip_fqdn_domain_pubsuffix_tld[0]
element_fqdn = element_ip_fqdn_domain_pubsuffix_tld[1]
element_domain = element_ip_fqdn_domain_pubsuffix_tld[2]
element_pubsuffix = element_ip_fqdn_domain_pubsuffix_tld[3]
element_tld = element_ip_fqdn_domain_pubsuffix_tld[4]
# sql_driver.add_domain both stores the new domain and returns its db row id
element_domain_id = sql_driver.add_domain(element_ip, element_fqdn,
element_domain,
element_pubsuffix,
element_tld)
# mark third-party elements based on domain
if origin_domain != element_domain:
is_3p_element = True
else:
is_3p_element = False
# if we are not storing 1p elements continue
if is_3p_element is False and store_1p is False:
continue
if request[:5] == 'https':
element_is_ssl = True
else:
element_is_ssl = False
try:
received = browser_output['processed_requests'][request][
'received']
except:
received = None
# get domain of referer and determine if page leaked by referer
try:
referer = browser_output['processed_requests'][request][
'referer']
except:
referer = None
if referer and len(referer) != 0:
referer_ip_fqdn_domain_pubsuffix_tld = self.url_parser.get_ip_fqdn_domain_pubsuffix_tld(
referer)
if referer_ip_fqdn_domain_pubsuffix_tld:
if referer_ip_fqdn_domain_pubsuffix_tld[
2] == origin_domain:
page_domain_in_referer = True
else:
page_domain_in_referer = False
else:
page_domain_in_referer = None
sql_driver.log_error(
url, 'Error parsing referer header: ' + referer)
else:
page_domain_in_referer = None
try:
start_time_offset = browser_output['processed_requests'][
request]['start_time_offset']
except:
start_time_offset = None
try:
load_time = browser_output['processed_requests'][request][
'load_time']
except:
load_time = None
try:
status = browser_output['processed_requests'][request][
'status']
except:
status = None
try:
status_text = browser_output['processed_requests'][request][
'status_text']
except:
status_text = None
try:
content_type = browser_output['processed_requests'][request][
'content_type']
except:
content_type = None
try:
body_size = browser_output['processed_requests'][request][
'body_size']
except:
body_size = None
try:
request_headers = str(browser_output['processed_requests']
[request]['request_headers'])
except:
request_headers = None
try:
response_headers = str(browser_output['processed_requests']
[request]['response_headers'])
except:
response_headers = None
# consider anything before the "?" to be the element_url
try:
element_url = re.search('^(.+?)\?.+$', request).group(1)
except:
element_url = request
# consider anything after the "?" to be the args
try:
element_args = re.search('^.+(\?.+)$',
request).group(1) # start url args
except:
element_args = None
# attempt to parse off the extension
try:
element_extension = re.search('\.([0-9A-Za-z]+)$',
element_url).group(1).lower()
except:
element_extension = None
# lists of common extensions, can be expanded
image_extensions = [
'png', 'jpg', 'jpgx', 'jpeg', 'gif', 'svg', 'bmp', 'tif',
'tiff', 'webp', 'srf'
]
script_extensions = ['js', 'javascript']
data_extensions = ['json', 'jsonp', 'xml']
font_extentions = ['woff', 'ttf', 'otf']
static_extentions = ['html', 'htm', 'shtml']
dynamic_extentions = [
'php', 'asp', 'jsp', 'aspx', 'ashx', 'pl', 'cgi', 'fcgi'
]
# figure out what type of element it is
if element_extension in image_extensions:
element_type = 'image'
elif element_extension in script_extensions:
element_type = 'javascript'
elif element_extension in data_extensions:
element_type = 'data_structured'
elif element_extension == 'css':
element_type = 'style_sheet'
elif element_extension in font_extentions:
element_type = 'font'
elif element_extension in static_extentions:
element_type = 'page_static'
elif element_extension == dynamic_extentions:
element_type = 'page_dynamic'
elif element_extension == 'swf' or element_extension == 'fla':
element_type = 'Shockwave Flash'
else:
element_type = None
# file hashing has non-trivial overhead and off by default
#
# what this does is uses the same ua/referer as the actual request
# so we are just replaying the last one to get similar response
# note that we aren't sending the same cookies so that could be an issue
# otherwise it is equivalent to a page refresh in theory
# option to hash only 3p elements observed here
if (get_file_hashes and hash_3p_only
and is_3p_element) or (get_file_hashes
and hash_3p_only == False):
replay_element_request = urllib.request.Request(
request,
headers={
'User-Agent':
browser_output['processed_requests'][request]
['user_agent'],
'Referer':
referer,
'Accept':
'*/*'
})
try:
file_md5 = hashlib.md5(
urllib.request.urlopen(replay_element_request,
timeout=10).read()).hexdigest()
except:
file_md5 = None
else:
file_md5 = None
# store request
sql_driver.add_element(
page_id, request, element_url, is_3p_element, element_is_ssl,
received, referer, page_domain_in_referer, start_time_offset,
load_time, status, status_text, content_type, body_size,
request_headers, response_headers, file_md5, element_extension,
element_type, element_args, element_domain_id)
# close db connection
sql_driver.close()
return True
