def crawl_worker(agent_cfg, url_tuple):
"""Crawl given url. Will work in parallel. Cannot be class method."""
MAX_SLEEP_BEFORE_JOB = 10 # prevent starting all parallel processes at the same instance
sleep(random() * MAX_SLEEP_BEFORE_JOB) # sleep for a while
try:
idx, url = url_tuple
idx = str(idx)
stdout_log = os.path.join(agent_cfg['job_dir'], fu.get_out_filename_from_url(url, str(idx), '.txt'))
if not url[:5] in ('data:', 'http:', 'https', 'file:'):
url = 'http://' + url
proxy_opt = mitm.init_mitmproxy(stdout_log[:-4], agent_cfg['timeout'], agent_cfg['mitm_proxy_logs']) if agent_cfg['use_mitm_proxy'] else ""
if not 'chrome_clicker' in agent_cfg['type']:
cmd = get_visit_cmd(agent_cfg, proxy_opt, stdout_log, url)
wl_log.info('>> %s (%s) %s' % (url, idx, cmd))
status, output = ut.run_cmd(cmd) # Run the command
if status and status != ERR_CMD_TIMEDOUT:
wl_log.critical('Error while visiting %s(%s) w/ command: %s: (%s) %s' % (url, idx, cmd, status, output))
else:
wl_log.info(' >> ok %s (%s)' % (url, idx))
else:
cr.crawl_url(agent_cfg['type'], url, proxy_opt)
sleep(2) # this will make sure mitmdump is timed out before we start to process the network dump
if agent_cfg['post_visit_func']: # this pluggable function will parse the logs and do whatever we want
agent_cfg['post_visit_func'](stdout_log, crawl_id=agent_cfg['crawl_id'], url=url)
except Exception as exc:
wl_log.exception('Exception in worker function %s %s' % (url_tuple, exc))
def _new_identity(self):
wl_log.info("Creating a new identity...")
try:
ActionChains(self.driver).send_keys(Keys.CONTROL + Keys.SHIFT +
'U').perform()
except WebDriverException:
pass
except Exception:
wl_log.exception("Exception while creating new identity.")
def add_captcha(self):
try:
captcha_filepath = ut.capture_dirpath_to_captcha(self.path)
move(self.path, captcha_filepath)
self.captchas[self.global_visit] = True
except OSError as e:
wl_log.exception('%s could not be renamed to %s', self.path,
captcha_filepath)
raise e
def crawl_worker(agent_cfg, url_tuple):
"""Crawl given url. Will work in parallel. Cannot be class method."""
MAX_SLEEP_BEFORE_JOB = 10 # prevent starting all parallel processes at the same instance
sleep(random() * MAX_SLEEP_BEFORE_JOB) # sleep for a while
try:
idx, url = url_tuple
idx = str(idx)
stdout_log = os.path.join(
agent_cfg['job_dir'],
fu.get_out_filename_from_url(url, str(idx), '.txt'))
if not url[:5] in ('data:', 'http:', 'https', 'file:'):
url = 'http://' + url
proxy_opt = mitm.init_mitmproxy(
stdout_log[:-4], agent_cfg['timeout'], agent_cfg['mitm_proxy_logs']
) if agent_cfg['use_mitm_proxy'] else ""
if not 'chrome_clicker' in agent_cfg['type']:
cmd = get_visit_cmd(agent_cfg, proxy_opt, stdout_log, url)
wl_log.info('>> %s (%s) %s' % (url, idx, cmd))
status, output = ut.run_cmd(cmd) # Run the command
if status and status != ERR_CMD_TIMEDOUT:
wl_log.critical(
'Error while visiting %s(%s) w/ command: %s: (%s) %s' %
(url, idx, cmd, status, output))
else:
wl_log.info(' >> ok %s (%s)' % (url, idx))
else:
cr.crawl_url(agent_cfg['type'], url, proxy_opt)
sleep(
2
) # this will make sure mitmdump is timed out before we start to process the network dump
if agent_cfg[
'post_visit_func']: # this pluggable function will parse the logs and do whatever we want
agent_cfg['post_visit_func'](stdout_log,
crawl_id=agent_cfg['crawl_id'],
url=url)
except Exception as exc:
wl_log.exception('Exception in worker function %s %s' %
(url_tuple, exc))
def parse_crawl_log(filename, dump_fun=None, crawl_id=0, url=""):
"""Populate domain info object by parsing crawl log file of a site.
Call dump function to output dump log.
Logs to be parsed with this function are generated by setting env. variable FC_DEBUG=1 to 1 or logs from the browser.
See, fontconfig library for details.
"""
origins_to_fonts = {} # will keep origin to loaded fonts mapping
domaInfo = DomainInfo()
file_content = fu.read_file(filename)
wl_log.info("Parsing log for %s %s" % (url, filename))
fonts_by_fc_debug = re.findall(
r"Sort Pattern.*$\W+family: \"([^\"]*)", file_content, re.MULTILINE
) # match family field of font request (not the matched one)
domaInfo.num_offsetWidth_calls = len(re.findall(r"Element::offsetWidth", file_content)) # offset width attempts
domaInfo.num_offsetHeight_calls = len(re.findall(r"Element::offsetHeight", file_content)) # offset height attempts
# TODO add getBoundingClientRect
font_and_urls = re.findall(
r"CSSFontSelector::getFontData:? (.*) ([^\s]*)", file_content
) # output from modified browser
# print 'font_and_urls', font_and_urls
font_face_pairs = re.findall(r"CSSFontFace::getFontData (.*)->(.*)", file_content) # output from modified browser
# print 'font_and_urls', font_and_urls
domaInfo.log_complete = int(bool(re.findall(r"Finished all steps!", file_content))) # output from modified browser
# print 'domaInfo.log_complete', domaInfo.log_complete
js_log_prefix = ">>>FPLOG"
fpd_logs = re.findall(r"%s.*" % js_log_prefix, file_content) # output from modified browser
domaInfo.fpd_logs = [call[len(js_log_prefix) + 1 :] for call in set(fpd_logs)]
for font_name, font_url in font_and_urls:
if font_url.startswith("http") and len(font_name) > 1 and not font_name[:5] in ("data:", "http:", "https"):
# font_name = font_name.rsplit(' ', 1)[0] if font_name.endswith(' onURL:') else font_name # TODO: unify chrome source code to log as Phantom do. then remove this line
font_name = font_name.lower().strip()
# origin = pub_suffix.get_public_suffix(font_url)\
origin = font_url
if origin in origins_to_fonts:
origins_to_fonts[origin].add(font_name)
# print 'added', font_name, 'to', origin, origins_to_fonts[origin]
else:
origins_to_fonts[origin] = set([font_name])
for font, face in font_face_pairs:
font = font.lower().strip()
face = face.lower().strip()
# replace all occurrences of this font-family name with the face
for fonts_by_origin in origins_to_fonts.itervalues():
try:
fonts_by_origin.remove(font)
except: # we cannot find this font in this origin's list
pass
else:
fonts_by_origin.add(face)
# print 'removed', font, 'added', face
for origin, fonts in origins_to_fonts.iteritems():
domaInfo.fonts_by_origins[origin] = list(fonts)
domaInfo.fonts_loaded += domaInfo.fonts_by_origins[origin]
domaInfo.fc_dbg_font_loads = list(
set([font.lower() for font in fonts_by_fc_debug if not font[:5] in ("data:", "http:", "https")])
) # filter out the data urls and web fonts
domaInfo.fonts_loaded = list(
set([font.lower() for font in domaInfo.fonts_loaded if not font[:5] in ("data:", "http:", "https")])
) # filter out the data urls and web fonts
requests = re.findall(r"^requested: (http.*)", file_content, re.MULTILINE)
if not requests and filename.endswith(MITM_LOG_EXTENSION):
requests = re.findall(r"(http.*)", file_content, re.MULTILINE)
responses = ""
# populate domain info obj
domaInfo.num_font_loads = len(domaInfo.fonts_loaded)
domaInfo.requests = list(set(requests))
domaInfo.responses = list(set(responses))
domaInfo.fp_detected = get_fp_from_reqs(requests)
domaInfo.url = url
domaInfo.rank = get_rank_domain_from_filename(filename)[
0
] # !!! rank may not be right. It's only true if we make top Alexa crawl.
domaInfo.log_filename = filename
domaInfo.crawl_id = crawl_id
# Read canvas events and print them to log in canvas
urls_read_from_canvas = Set()
urls_wrote_to_canvas = Set()
canvas_log = os.path.join(cm.BASE_FP_LOGS_FOLDER, str(crawl_id) + "canvas.log")
read = wrote = False
for read_event in cm.CANVAS_READ_EVENTS:
if read_event in file_content:
read = True
break
for write_event in cm.CANVAS_WRITE_EVENTS:
if write_event in file_content:
wrote = True
break
if read and wrote:
wl_log.info("Found both canvas read and write events in log %s, registering in : %s" % (filename, canvas_log))
with open(canvas_log, "a+") as f:
f.write(" ".join([str(domaInfo.rank), domaInfo.url]) + "\n")
if dump_fun: # call dump function
try:
dump_fun(domaInfo)
except KeyboardInterrupt:
raise
except Exception as exc:
wl_log.exception("Exception while dumping %s: %s" % (domaInfo.url, exc))
def _do_visits(self):
self._config_driver()
with TorBrowserDriver(**self.driver_config) as driver:
self.driver = driver
failed = 0
while self.job.visit < self.job.visits:
self.job.visit += 1
raised = False
num_retry, retry = 0, RETRY
while True:
wl_log.info("*** Visit #%s to %s ***", self.job.visit,
self.job.url)
try:
ut.create_dir(self.job.path)
self.save_checkpoint()
self.set_page_load_timeout()
try:
self._do_instance()
self.get_screenshot_if_enabled()
except (cm.HardTimeoutException, TimeoutException,
SnifferTimeoutError):
wl_log.exception("Visit to %s has timed out!",
self.job.url)
else:
self.post_visit()
finally:
self._reset_tor()
self.cleanup_visit()
except OSError as ose:
wl_log.exception("OS Error %s" % repr(ose))
raise ose # we better halt here, we may be out of disk space
except TBDriverPortError as tbde:
raise tbde # cannot connect to Tor, the following ones will fail
except cm.ConnErrorPage as cnpe:
raised = cnpe
except stem.SocketError as sckte:
raised = sckte
except Exception as exc:
wl_log.exception("Unknown exception: %s" % repr(exc))
raised = exc
finally:
if not raised:
break
# there was a non-timeout exception
failed += 1
if failed >= MAX_FAILED:
self._mark_failed()
self.job.visit = self.job.visits
break
# retry visit?
if not retry:
break
wl_log.info("Will retry the visit. Retry num %s/%s" %
(retry, cm.MAX_RETRIES))
rmtree(self.job.path)
num_retry += 1
if num_retry == cm.MAX_RETRIES:
retry = False
self.controller.restart_tor()
if self.job.visit == self.job.visits:
self.job.visit = 0
def parse_crawl_log(filename, dump_fun=None, crawl_id=0, url=""):
"""Populate domain info object by parsing crawl log file of a site.
Call dump function to output dump log.
Logs to be parsed with this function are generated by setting env. variable FC_DEBUG=1 to 1 or logs from the browser.
See, fontconfig library for details.
"""
origins_to_fonts = {} # will keep origin to loaded fonts mapping
domaInfo = DomainInfo()
file_content = fu.read_file(filename)
wl_log.info('Parsing log for %s %s' % (url, filename))
fonts_by_fc_debug = re.findall(
r"Sort Pattern.*$\W+family: \"([^\"]*)", file_content, re.MULTILINE
) # match family field of font request (not the matched one)
domaInfo.num_offsetWidth_calls = len(
re.findall(r"Element::offsetWidth",
file_content)) # offset width attempts
domaInfo.num_offsetHeight_calls = len(
re.findall(r"Element::offsetHeight",
file_content)) # offset height attempts
# TODO add getBoundingClientRect
font_and_urls = re.findall(r"CSSFontSelector::getFontData:? (.*) ([^\s]*)",
file_content) # output from modified browser
#print 'font_and_urls', font_and_urls
font_face_pairs = re.findall(r"CSSFontFace::getFontData (.*)->(.*)",
file_content) # output from modified browser
#print 'font_and_urls', font_and_urls
domaInfo.log_complete = int(
bool(re.findall(r"Finished all steps!",
file_content))) # output from modified browser
#print 'domaInfo.log_complete', domaInfo.log_complete
js_log_prefix = ">>>FPLOG"
fpd_logs = re.findall(r'%s.*' % js_log_prefix,
file_content) # output from modified browser
domaInfo.fpd_logs = [
call[len(js_log_prefix) + 1:] for call in set(fpd_logs)
]
for font_name, font_url in font_and_urls:
if font_url.startswith('http') and len(
font_name) > 1 and not font_name[:5] in ('data:', 'http:',
'https'):
#font_name = font_name.rsplit(' ', 1)[0] if font_name.endswith(' onURL:') else font_name # TODO: unify chrome source code to log as Phantom do. then remove this line
font_name = font_name.lower().strip()
# origin = pub_suffix.get_public_suffix(font_url)\
origin = font_url
if origin in origins_to_fonts:
origins_to_fonts[origin].add(font_name)
#print 'added', font_name, 'to', origin, origins_to_fonts[origin]
else:
origins_to_fonts[origin] = set([
font_name,
])
for font, face in font_face_pairs:
font = font.lower().strip()
face = face.lower().strip()
# replace all occurrences of this font-family name with the face
for fonts_by_origin in origins_to_fonts.itervalues():
try:
fonts_by_origin.remove(font)
except: # we cannot find this font in this origin's list
pass
else:
fonts_by_origin.add(face)
# print 'removed', font, 'added', face
for origin, fonts in origins_to_fonts.iteritems():
domaInfo.fonts_by_origins[origin] = list(fonts)
domaInfo.fonts_loaded += domaInfo.fonts_by_origins[origin]
domaInfo.fc_dbg_font_loads = list(set([font.lower() for font in fonts_by_fc_debug \
if not font[:5] in ('data:', 'http:', 'https')])) # filter out the data urls and web fonts
domaInfo.fonts_loaded = list(set([font.lower() for font in domaInfo.fonts_loaded \
if not font[:5] in ('data:', 'http:', 'https')])) # filter out the data urls and web fonts
requests = re.findall(r"^requested: (http.*)", file_content, re.MULTILINE)
if not requests and filename.endswith(MITM_LOG_EXTENSION):
requests = re.findall(r"(http.*)", file_content, re.MULTILINE)
responses = ''
# populate domain info obj
domaInfo.num_font_loads = len(domaInfo.fonts_loaded)
domaInfo.requests = list(set(requests))
domaInfo.responses = list(set(responses))
domaInfo.fp_detected = get_fp_from_reqs(requests)
domaInfo.url = url
domaInfo.rank = get_rank_domain_from_filename(
filename
)[0] # !!! rank may not be right. It's only true if we make top Alexa crawl.
domaInfo.log_filename = filename
domaInfo.crawl_id = crawl_id
# Read canvas events and print them to log in canvas
urls_read_from_canvas = Set()
urls_wrote_to_canvas = Set()
canvas_log = os.path.join(cm.BASE_FP_LOGS_FOLDER,
str(crawl_id) + "canvas.log")
read = wrote = False
for read_event in cm.CANVAS_READ_EVENTS:
if read_event in file_content:
read = True
break
for write_event in cm.CANVAS_WRITE_EVENTS:
if write_event in file_content:
wrote = True
break
if read and wrote:
wl_log.info(
'Found both canvas read and write events in log %s, registering in : %s'
% (filename, canvas_log))
with open(canvas_log, "a+") as f:
f.write(" ".join([str(domaInfo.rank), domaInfo.url]) + "\n")
if dump_fun: # call dump function
try:
dump_fun(domaInfo)
except KeyboardInterrupt:
raise
except Exception as exc:
wl_log.exception("Exception while dumping %s: %s" %
(domaInfo.url, exc))
