AIL-framework/var/www/modules/terms/Flask_terms.py

604 lines
24 KiB
Python

#!/usr/bin/env python2
# -*-coding:UTF-8 -*
'''
Flask functions and routes for the trending modules page
note: The matching of credential against supplied credential is done using Levenshtein distance
'''
import redis
import datetime
import calendar
import flask
from flask import Flask, render_template, jsonify, request, Blueprint
import re
import Paste
from pprint import pprint
import Levenshtein
# ============ VARIABLES ============
import Flask_config
app = Flask_config.app
cfg = Flask_config.cfg
r_serv_term = Flask_config.r_serv_term
r_serv_cred = Flask_config.r_serv_cred
terms = Blueprint('terms', __name__, template_folder='templates')
'''TERM'''
DEFAULT_MATCH_PERCENT = 50
#tracked
TrackedTermsSet_Name = "TrackedSetTermSet"
TrackedTermsDate_Name = "TrackedTermDate"
#black
BlackListTermsDate_Name = "BlackListTermDate"
BlackListTermsSet_Name = "BlackListSetTermSet"
#regex
TrackedRegexSet_Name = "TrackedRegexSet"
TrackedRegexDate_Name = "TrackedRegexDate"
#set
TrackedSetSet_Name = "TrackedSetSet"
TrackedSetDate_Name = "TrackedSetDate"
# notifications enabled/disabled
# same value as in `bin/NotificationHelper.py`
TrackedTermsNotificationEnabled_Name = "TrackedNotifications"
# associated notification email addresses for a specific term`
# same value as in `bin/NotificationHelper.py`
# Keys will be e.g. TrackedNotificationEmails_<TERMNAME>
TrackedTermsNotificationEmailsPrefix_Name = "TrackedNotificationEmails_"
'''CRED'''
REGEX_CRED = '[a-z]+|[A-Z]{3,}|[A-Z]{1,2}[a-z]+|[0-9]+'
REDIS_KEY_NUM_USERNAME = 'uniqNumForUsername'
REDIS_KEY_NUM_PATH = 'uniqNumForUsername'
REDIS_KEY_ALL_CRED_SET = 'AllCredentials'
REDIS_KEY_ALL_CRED_SET_REV = 'AllCredentialsRev'
REDIS_KEY_ALL_PATH_SET = 'AllPath'
REDIS_KEY_ALL_PATH_SET_REV = 'AllPathRev'
REDIS_KEY_MAP_CRED_TO_PATH = 'CredToPathMapping'
# ============ FUNCTIONS ============
def Term_getValueOverRange(word, startDate, num_day, per_paste=""):
passed_days = 0
oneDay = 60*60*24
to_return = []
curr_to_return = 0
for timestamp in range(startDate, startDate - max(num_day)*oneDay, -oneDay):
value = r_serv_term.hget(per_paste+str(timestamp), word)
curr_to_return += int(value) if value is not None else 0
for i in num_day:
if passed_days == i-1:
to_return.append(curr_to_return)
passed_days += 1
return to_return
#Mix suplied username, if extensive is set, slice username(s) with different windows
def mixUserName(supplied, extensive=False):
#e.g.: John Smith
terms = supplied.split()[:2]
usernames = []
if len(terms) == 1:
terms.append(' ')
#john, smith, John, Smith, JOHN, SMITH
usernames += [terms[0].lower()]
usernames += [terms[1].lower()]
usernames += [terms[0][0].upper() + terms[0][1:].lower()]
usernames += [terms[1][0].upper() + terms[1][1:].lower()]
usernames += [terms[0].upper()]
usernames += [terms[1].upper()]
#johnsmith, smithjohn, JOHNsmith, johnSMITH, SMITHjohn, smithJOHN
usernames += [(terms[0].lower() + terms[1].lower()).strip()]
usernames += [(terms[1].lower() + terms[0].lower()).strip()]
usernames += [(terms[0].upper() + terms[1].lower()).strip()]
usernames += [(terms[0].lower() + terms[1].upper()).strip()]
usernames += [(terms[1].upper() + terms[0].lower()).strip()]
usernames += [(terms[1].lower() + terms[0].upper()).strip()]
#Jsmith, JSmith, jsmith, jSmith, johnS, Js, JohnSmith, Johnsmith, johnSmith
usernames += [(terms[0][0].upper() + terms[1][0].lower() + terms[1][1:].lower()).strip()]
usernames += [(terms[0][0].upper() + terms[1][0].upper() + terms[1][1:].lower()).strip()]
usernames += [(terms[0][0].lower() + terms[1][0].lower() + terms[1][1:].lower()).strip()]
usernames += [(terms[0][0].lower() + terms[1][0].upper() + terms[1][1:].lower()).strip()]
usernames += [(terms[0].lower() + terms[1][0].upper()).strip()]
usernames += [(terms[0].upper() + terms[1][0].lower()).strip()]
usernames += [(terms[0][0].upper() + terms[0][1:].lower() + terms[1][0].upper() + terms[1][1:].lower()).strip()]
usernames += [(terms[0][0].upper() + terms[0][1:].lower() + terms[1][0].lower() + terms[1][1:].lower()).strip()]
usernames += [(terms[0][0].lower() + terms[0][1:].lower() + terms[1][0].upper() + terms[1][1:].lower()).strip()]
if not extensive:
return usernames
#Slice the supplied username(s)
mixedSupplied = supplied.replace(' ','')
minWindow = 3 if len(mixedSupplied)/2 < 4 else len(mixedSupplied)/2
for winSize in range(3,len(mixedSupplied)):
for startIndex in range(0, len(mixedSupplied)-winSize):
usernames += [mixedSupplied[startIndex:startIndex+winSize]]
filtered_usernames = []
for usr in usernames:
if len(usr) > 2:
filtered_usernames.append(usr)
return filtered_usernames
# ============ ROUTES ============
@terms.route("/terms_management/")
def terms_management():
per_paste = request.args.get('per_paste')
if per_paste == "1" or per_paste is None:
per_paste_text = "per_paste_"
per_paste = 1
else:
per_paste_text = ""
per_paste = 0
today = datetime.datetime.now()
today = today.replace(hour=0, minute=0, second=0, microsecond=0)
today_timestamp = calendar.timegm(today.timetuple())
# Map tracking if notifications are enabled for a specific term
notificationEnabledDict = {}
# Maps a specific term to the associated email addresses
notificationEMailTermMapping = {}
#Regex
trackReg_list = []
trackReg_list_values = []
trackReg_list_num_of_paste = []
for tracked_regex in r_serv_term.smembers(TrackedRegexSet_Name):
notificationEMailTermMapping[tracked_regex] = "\n".join(r_serv_term.smembers(TrackedTermsNotificationEmailsPrefix_Name + tracked_regex))
if tracked_regex not in notificationEnabledDict:
notificationEnabledDict[tracked_regex] = False
trackReg_list.append(tracked_regex)
value_range = Term_getValueOverRange(tracked_regex, today_timestamp, [1, 7, 31], per_paste=per_paste_text)
term_date = r_serv_term.hget(TrackedRegexDate_Name, tracked_regex)
set_paste_name = "regex_" + tracked_regex
trackReg_list_num_of_paste.append(r_serv_term.scard(set_paste_name))
term_date = datetime.datetime.utcfromtimestamp(int(term_date)) if term_date is not None else "No date recorded"
value_range.append(term_date)
trackReg_list_values.append(value_range)
if tracked_regex in r_serv_term.smembers(TrackedTermsNotificationEnabled_Name):
notificationEnabledDict[tracked_regex] = True
#Set
trackSet_list = []
trackSet_list_values = []
trackSet_list_num_of_paste = []
for tracked_set in r_serv_term.smembers(TrackedSetSet_Name):
notificationEMailTermMapping[tracked_set] = "\n".join(r_serv_term.smembers(TrackedTermsNotificationEmailsPrefix_Name + tracked_set))
if tracked_set not in notificationEnabledDict:
notificationEnabledDict[tracked_set] = False
trackSet_list.append(tracked_set)
value_range = Term_getValueOverRange(tracked_set, today_timestamp, [1, 7, 31], per_paste=per_paste_text)
term_date = r_serv_term.hget(TrackedSetDate_Name, tracked_set)
set_paste_name = "set_" + tracked_set
trackSet_list_num_of_paste.append(r_serv_term.scard(set_paste_name))
term_date = datetime.datetime.utcfromtimestamp(int(term_date)) if term_date is not None else "No date recorded"
value_range.append(term_date)
trackSet_list_values.append(value_range)
if tracked_set in r_serv_term.smembers(TrackedTermsNotificationEnabled_Name):
notificationEnabledDict[tracked_set] = True
#Tracked terms
track_list = []
track_list_values = []
track_list_num_of_paste = []
for tracked_term in r_serv_term.smembers(TrackedTermsSet_Name):
notificationEMailTermMapping[tracked_term] = "\n".join(r_serv_term.smembers(TrackedTermsNotificationEmailsPrefix_Name + tracked_term))
if tracked_term not in notificationEnabledDict:
notificationEnabledDict[tracked_term] = False
track_list.append(tracked_term)
value_range = Term_getValueOverRange(tracked_term, today_timestamp, [1, 7, 31], per_paste=per_paste_text)
term_date = r_serv_term.hget(TrackedTermsDate_Name, tracked_term)
set_paste_name = "tracked_" + tracked_term
track_list_num_of_paste.append(r_serv_term.scard(set_paste_name))
term_date = datetime.datetime.utcfromtimestamp(int(term_date)) if term_date is not None else "No date recorded"
value_range.append(term_date)
track_list_values.append(value_range)
if tracked_term in r_serv_term.smembers(TrackedTermsNotificationEnabled_Name):
notificationEnabledDict[tracked_term] = True
#blacklist terms
black_list = []
for blacked_term in r_serv_term.smembers(BlackListTermsSet_Name):
term_date = r_serv_term.hget(BlackListTermsDate_Name, blacked_term)
term_date = datetime.datetime.utcfromtimestamp(int(term_date)) if term_date is not None else "No date recorded"
black_list.append([blacked_term, term_date])
return render_template("terms_management.html",
black_list=black_list, track_list=track_list, trackReg_list=trackReg_list, trackSet_list=trackSet_list,
track_list_values=track_list_values, track_list_num_of_paste=track_list_num_of_paste,
trackReg_list_values=trackReg_list_values, trackReg_list_num_of_paste=trackReg_list_num_of_paste,
trackSet_list_values=trackSet_list_values, trackSet_list_num_of_paste=trackSet_list_num_of_paste,
per_paste=per_paste, notificationEnabledDict=notificationEnabledDict, notificationEMailTermMapping=notificationEMailTermMapping)
@terms.route("/terms_management_query_paste/")
def terms_management_query_paste():
term = request.args.get('term')
paste_info = []
# check if regex or not
if term.startswith('/') and term.endswith('/'):
set_paste_name = "regex_" + term
track_list_path = r_serv_term.smembers(set_paste_name)
elif term.startswith('\\') and term.endswith('\\'):
set_paste_name = "set_" + term
track_list_path = r_serv_term.smembers(set_paste_name)
else:
set_paste_name = "tracked_" + term
track_list_path = r_serv_term.smembers(set_paste_name)
for path in track_list_path:
paste = Paste.Paste(path)
p_date = str(paste._get_p_date())
p_date = p_date[6:]+'/'+p_date[4:6]+'/'+p_date[0:4]
p_source = paste.p_source
p_encoding = paste._get_p_encoding()
p_size = paste.p_size
p_mime = paste.p_mime
p_lineinfo = paste.get_lines_info()
p_content = paste.get_p_content().decode('utf-8', 'ignore')
if p_content != 0:
p_content = p_content[0:400]
paste_info.append({"path": path, "date": p_date, "source": p_source, "encoding": p_encoding, "size": p_size, "mime": p_mime, "lineinfo": p_lineinfo, "content": p_content})
return jsonify(paste_info)
@terms.route("/terms_management_query/")
def terms_management_query():
TrackedTermsDate_Name = "TrackedTermDate"
BlackListTermsDate_Name = "BlackListTermDate"
term = request.args.get('term')
section = request.args.get('section')
today = datetime.datetime.now()
today = today.replace(hour=0, minute=0, second=0, microsecond=0)
today_timestamp = calendar.timegm(today.timetuple())
value_range = Term_getValueOverRange(term, today_timestamp, [1, 7, 31])
if section == "followTerm":
term_date = r_serv_term.hget(TrackedTermsDate_Name, term)
elif section == "blacklistTerm":
term_date = r_serv_term.hget(BlackListTermsDate_Name, term)
term_date = datetime.datetime.utcfromtimestamp(int(term_date)) if term_date is not None else "No date recorded"
value_range.append(str(term_date))
return jsonify(value_range)
@terms.route("/terms_management_action/", methods=['GET'])
def terms_management_action():
today = datetime.datetime.now()
today = today.replace(microsecond=0)
today_timestamp = calendar.timegm(today.timetuple())
section = request.args.get('section')
action = request.args.get('action')
term = request.args.get('term')
notificationEmailsParam = request.args.get('emailAddresses')
if action is None or term is None:
return "None"
else:
if section == "followTerm":
if action == "add":
# Strip all whitespace
notificationEmailsParam = "".join(notificationEmailsParam.split())
# Make a list of all passed email addresses
notificationEmails = notificationEmailsParam.split(",")
validNotificationEmails = []
# check for valid email addresses
for email in notificationEmails:
# Really basic validation:
# has exactly one @ sign, and at least one . in the part after the @
if re.match(r"[^@]+@[^@]+\.[^@]+", email):
validNotificationEmails.append(email)
# check if regex/set or simple term
#regex
if term.startswith('/') and term.endswith('/'):
r_serv_term.sadd(TrackedRegexSet_Name, term)
r_serv_term.hset(TrackedRegexDate_Name, term, today_timestamp)
# add all valid emails to the set
for email in validNotificationEmails:
r_serv_term.sadd(TrackedTermsNotificationEmailsPrefix_Name + term, email)
# enable notifications by default
r_serv_term.sadd(TrackedTermsNotificationEnabled_Name, term)
#set
elif term.startswith('\\') and term.endswith('\\'):
tab_term = term[1:-1]
perc_finder = re.compile("\[[0-9]{1,3}\]").search(tab_term)
if perc_finder is not None:
match_percent = perc_finder.group(0)[1:-1]
set_to_add = term
else:
match_percent = DEFAULT_MATCH_PERCENT
set_to_add = "\\" + tab_term[:-1] + ", [{}]]\\".format(match_percent)
r_serv_term.sadd(TrackedSetSet_Name, set_to_add)
r_serv_term.hset(TrackedSetDate_Name, set_to_add, today_timestamp)
# add all valid emails to the set
for email in validNotificationEmails:
r_serv_term.sadd(TrackedTermsNotificationEmailsPrefix_Name + set_to_add, email)
# enable notifications by default
r_serv_term.sadd(TrackedTermsNotificationEnabled_Name, set_to_add)
#simple term
else:
r_serv_term.sadd(TrackedTermsSet_Name, term.lower())
r_serv_term.hset(TrackedTermsDate_Name, term.lower(), today_timestamp)
# add all valid emails to the set
for email in validNotificationEmails:
r_serv_term.sadd(TrackedTermsNotificationEmailsPrefix_Name + term.lower(), email)
# enable notifications by default
r_serv_term.sadd(TrackedTermsNotificationEnabled_Name, term.lower())
elif action == "toggleEMailNotification":
# get the current state
if term in r_serv_term.smembers(TrackedTermsNotificationEnabled_Name):
# remove it
r_serv_term.srem(TrackedTermsNotificationEnabled_Name, term.lower())
else:
# add it
r_serv_term.sadd(TrackedTermsNotificationEnabled_Name, term.lower())
#del action
else:
if term.startswith('/') and term.endswith('/'):
r_serv_term.srem(TrackedRegexSet_Name, term)
r_serv_term.hdel(TrackedRegexDate_Name, term)
elif term.startswith('\\') and term.endswith('\\'):
r_serv_term.srem(TrackedSetSet_Name, term)
print(term)
r_serv_term.hdel(TrackedSetDate_Name, term)
else:
r_serv_term.srem(TrackedTermsSet_Name, term.lower())
r_serv_term.hdel(TrackedTermsDate_Name, term.lower())
# delete the associated notification emails too
r_serv_term.delete(TrackedTermsNotificationEmailsPrefix_Name + term)
elif section == "blacklistTerm":
if action == "add":
r_serv_term.sadd(BlackListTermsSet_Name, term.lower())
r_serv_term.hset(BlackListTermsDate_Name, term, today_timestamp)
else:
r_serv_term.srem(BlackListTermsSet_Name, term.lower())
else:
return "None"
to_return = {}
to_return["section"] = section
to_return["action"] = action
to_return["term"] = term
return jsonify(to_return)
@terms.route("/terms_plot_tool/")
def terms_plot_tool():
term = request.args.get('term')
if term is not None:
return render_template("terms_plot_tool.html", term=term)
else:
return render_template("terms_plot_tool.html", term="")
@terms.route("/terms_plot_tool_data/")
def terms_plot_tool_data():
oneDay = 60*60*24
range_start = datetime.datetime.utcfromtimestamp(int(float(request.args.get('range_start')))) if request.args.get('range_start') is not None else 0;
range_start = range_start.replace(hour=0, minute=0, second=0, microsecond=0)
range_start = calendar.timegm(range_start.timetuple())
range_end = datetime.datetime.utcfromtimestamp(int(float(request.args.get('range_end')))) if request.args.get('range_end') is not None else 0;
range_end = range_end.replace(hour=0, minute=0, second=0, microsecond=0)
range_end = calendar.timegm(range_end.timetuple())
term = request.args.get('term')
per_paste = request.args.get('per_paste')
if per_paste == "1" or per_paste is None:
per_paste = "per_paste_"
else:
per_paste = ""
if term is None:
return "None"
else:
value_range = []
for timestamp in range(range_start, range_end+oneDay, oneDay):
value = r_serv_term.hget(per_paste+str(timestamp), term)
curr_value_range = int(value) if value is not None else 0
value_range.append([timestamp, curr_value_range])
value_range.insert(0,term)
return jsonify(value_range)
@terms.route("/terms_plot_top/")
def terms_plot_top():
per_paste = request.args.get('per_paste')
per_paste = per_paste if per_paste is not None else 1
return render_template("terms_plot_top.html", per_paste=per_paste)
@terms.route("/terms_plot_top_data/")
def terms_plot_top_data():
oneDay = 60*60*24
today = datetime.datetime.now()
today = today.replace(hour=0, minute=0, second=0, microsecond=0)
today_timestamp = calendar.timegm(today.timetuple())
per_paste = request.args.get('per_paste')
if per_paste == "1" or per_paste is None:
per_paste = "per_paste_"
else:
per_paste = ""
set_day = per_paste + "TopTermFreq_set_day_" + str(today_timestamp)
set_week = per_paste + "TopTermFreq_set_week";
set_month = per_paste + "TopTermFreq_set_month";
the_set = per_paste + request.args.get('set')
num_day = int(request.args.get('num_day'))
if the_set is None:
return "None"
else:
to_return = []
if "TopTermFreq_set_day" in the_set:
the_set += "_" + str(today_timestamp)
for term, tot_value in r_serv_term.zrevrangebyscore(the_set, '+inf', '-inf', withscores=True, start=0, num=20):
position = {}
position['day'] = r_serv_term.zrevrank(set_day, term)
position['day'] = position['day']+1 if position['day'] is not None else "<20"
position['week'] = r_serv_term.zrevrank(set_week, term)
position['week'] = position['week']+1 if position['week'] is not None else "<20"
position['month'] = r_serv_term.zrevrank(set_month, term)
position['month'] = position['month']+1 if position['month'] is not None else "<20"
value_range = []
for timestamp in range(today_timestamp, today_timestamp - num_day*oneDay, -oneDay):
value = r_serv_term.hget(per_paste+str(timestamp), term)
curr_value_range = int(value) if value is not None else 0
value_range.append([timestamp, curr_value_range])
to_return.append([term, value_range, tot_value, position])
return jsonify(to_return)
@terms.route("/credentials_tracker/")
def credentials_tracker():
return render_template("credentials_tracker.html")
@terms.route("/credentials_management_query_paste/", methods=['GET', 'POST'])
def credentials_management_query_paste():
cred = request.args.get('cred')
allPath = request.json['allPath']
paste_info = []
for pathNum in allPath:
path = r_serv_cred.hget(REDIS_KEY_ALL_PATH_SET_REV, pathNum)
paste = Paste.Paste(path)
p_date = str(paste._get_p_date())
p_date = p_date[6:]+'/'+p_date[4:6]+'/'+p_date[0:4]
p_source = paste.p_source
p_encoding = paste._get_p_encoding()
p_size = paste.p_size
p_mime = paste.p_mime
p_lineinfo = paste.get_lines_info()
p_content = paste.get_p_content().decode('utf-8', 'ignore')
if p_content != 0:
p_content = p_content[0:400]
paste_info.append({"path": path, "date": p_date, "source": p_source, "encoding": p_encoding, "size": p_size, "mime": p_mime, "lineinfo": p_lineinfo, "content": p_content})
return jsonify(paste_info)
@terms.route("/credentials_management_action/", methods=['GET'])
def cred_management_action():
supplied = request.args.get('term').encode('utf-8')
action = request.args.get('action')
section = request.args.get('section')
extensive = request.args.get('extensive')
extensive = True if extensive == "true" else False
if extensive:
#collectDico
AllUsernameInRedis = r_serv_cred.hgetall(REDIS_KEY_ALL_CRED_SET).keys()
uniq_num_set = set()
if action == "seek":
possibilities = mixUserName(supplied, extensive)
for poss in possibilities:
num = r_serv_cred.hget(REDIS_KEY_ALL_CRED_SET, poss)
if num is not None:
uniq_num_set.add(num)
for num in r_serv_cred.smembers(poss):
uniq_num_set.add(num)
#Extensive /!\
if extensive:
iter_num = 0
tot_iter = len(AllUsernameInRedis)*len(possibilities)
for tempUsername in AllUsernameInRedis:
for poss in possibilities:
#FIXME print progress
if(iter_num % int(tot_iter/20) == 0):
#print("searching: {}% done".format(int(iter_num/tot_iter*100)), sep=' ', end='\r', flush=True)
print("searching: {}% done".format(float(iter_num)/float(tot_iter)*100))
iter_num += 1
if poss in tempUsername:
num = r_serv_cred.hget(REDIS_KEY_ALL_CRED_SET, tempUsername)
if num is not None:
uniq_num_set.add(num)
for num in r_serv_cred.smembers(tempUsername):
uniq_num_set.add(num)
data = {'usr': [], 'path': [], 'numPaste': [], 'simil': []}
for Unum in uniq_num_set:
levenRatio = 2.0
username = r_serv_cred.hget(REDIS_KEY_ALL_CRED_SET_REV, Unum)
# Calculate Levenshtein distance, ignore negative ratio
supp_splitted = supplied.split()
supp_mixed = supplied.replace(' ','')
supp_splitted.append(supp_mixed)
for indiv_supplied in supp_splitted:
levenRatio = float(Levenshtein.ratio(indiv_supplied, username))
levenRatioStr = "{:.1%}".format(levenRatio)
data['usr'].append(username)
allPathNum = list(r_serv_cred.smembers(REDIS_KEY_MAP_CRED_TO_PATH+'_'+Unum))
data['path'].append(allPathNum)
data['numPaste'].append(len(allPathNum))
data['simil'].append(levenRatioStr)
to_return = {}
to_return["section"] = section
to_return["action"] = action
to_return["term"] = supplied
to_return["data"] = data
return jsonify(to_return)
# ========= REGISTRATION =========
app.register_blueprint(terms)