Created
November 14, 2015 12:33
-
-
Save wienczny/a22256bb240d0492d126 to your computer and use it in GitHub Desktop.
Python3 pwdhash.py for https://www.pwdhash.com/
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#!/usr/bin/env python | |
import base64 | |
import re | |
import hashlib | |
import hmac | |
import itertools | |
def b64_hmac_md5(key, data): | |
""" | |
return base64-encoded HMAC-MD5 for key and data, with trailing '=' | |
stripped. | |
""" | |
digest = hmac.new(key.encode('utf-8'), data.encode('utf-8'), hashlib.md5).digest() | |
bdigest = base64.b64encode(digest).strip() | |
return re.sub('=+$', '', bdigest.decode('utf-8')) | |
# set of domain suffixes to be kept | |
_domains = ["ab.ca", "ac.ac", "ac.at", "ac.be", "ac.cn", "ac.il", | |
"ac.in", "ac.jp", "ac.kr", "ac.nz", "ac.th", "ac.uk", | |
"ac.za", "adm.br", "adv.br", "agro.pl", "ah.cn", "aid.pl", | |
"alt.za", "am.br", "arq.br", "art.br", "arts.ro", | |
"asn.au", "asso.fr", "asso.mc", "atm.pl", "auto.pl", | |
"bbs.tr", "bc.ca", "bio.br", "biz.pl", "bj.cn", "br.com", | |
"cn.com", "cng.br", "cnt.br", "co.ac", "co.at", "co.il", | |
"co.in", "co.jp", "co.kr", "co.nz", "co.th", "co.uk", | |
"co.za", "com.au", "com.br", "com.cn", "com.ec", "com.fr", | |
"com.hk", "com.mm", "com.mx", "com.pl", "com.ro", | |
"com.ru", "com.sg", "com.tr", "com.tw", "cq.cn", "cri.nz", | |
"de.com", "ecn.br", "edu.au", "edu.cn", "edu.hk", | |
"edu.mm", "edu.mx", "edu.pl", "edu.tr", "edu.za", | |
"eng.br", "ernet.in", "esp.br", "etc.br", "eti.br", | |
"eu.com", "eu.lv", "fin.ec", "firm.ro", "fm.br", "fot.br", | |
"fst.br", "g12.br", "gb.com", "gb.net", "gd.cn", "gen.nz", | |
"gmina.pl", "go.jp", "go.kr", "go.th", "gob.mx", "gov.br", | |
"gov.cn", "gov.ec", "gov.il", "gov.in", "gov.mm", | |
"gov.mx", "gov.sg", "gov.tr", "gov.za", "govt.nz", | |
"gs.cn", "gsm.pl", "gv.ac", "gv.at", "gx.cn", "gz.cn", | |
"hb.cn", "he.cn", "hi.cn", "hk.cn", "hl.cn", "hn.cn", | |
"hu.com", "idv.tw", "ind.br", "inf.br", "info.pl", | |
"info.ro", "iwi.nz", "jl.cn", "jor.br", "jpn.com", | |
"js.cn", "k12.il", "k12.tr", "lel.br", "ln.cn", "ltd.uk", | |
"mail.pl", "maori.nz", "mb.ca", "me.uk", "med.br", | |
"med.ec", "media.pl", "mi.th", "miasta.pl", "mil.br", | |
"mil.ec", "mil.nz", "mil.pl", "mil.tr", "mil.za", "mo.cn", | |
"muni.il", "nb.ca", "ne.jp", "ne.kr", "net.au", "net.br", | |
"net.cn", "net.ec", "net.hk", "net.il", "net.in", | |
"net.mm", "net.mx", "net.nz", "net.pl", "net.ru", | |
"net.sg", "net.th", "net.tr", "net.tw", "net.za", "nf.ca", | |
"ngo.za", "nm.cn", "nm.kr", "no.com", "nom.br", "nom.pl", | |
"nom.ro", "nom.za", "ns.ca", "nt.ca", "nt.ro", "ntr.br", | |
"nx.cn", "odo.br", "on.ca", "or.ac", "or.at", "or.jp", | |
"or.kr", "or.th", "org.au", "org.br", "org.cn", "org.ec", | |
"org.hk", "org.il", "org.mm", "org.mx", "org.nz", | |
"org.pl", "org.ro", "org.ru", "org.sg", "org.tr", | |
"org.tw", "org.uk", "org.za", "pc.pl", "pe.ca", "plc.uk", | |
"ppg.br", "presse.fr", "priv.pl", "pro.br", "psc.br", | |
"psi.br", "qc.ca", "qc.com", "qh.cn", "re.kr", | |
"realestate.pl", "rec.br", "rec.ro", "rel.pl", "res.in", | |
"ru.com", "sa.com", "sc.cn", "school.nz", "school.za", | |
"se.com", "se.net", "sh.cn", "shop.pl", "sk.ca", | |
"sklep.pl", "slg.br", "sn.cn", "sos.pl", "store.ro", | |
"targi.pl", "tj.cn", "tm.fr", "tm.mc", "tm.pl", "tm.ro", | |
"tm.za", "tmp.br", "tourism.pl", "travel.pl", "tur.br", | |
"turystyka.pl", "tv.br", "tw.cn", "uk.co", "uk.com", | |
"uk.net", "us.com", "uy.com", "vet.br", "web.za", | |
"web.com", "www.ro", "xj.cn", "xz.cn", "yk.ca", "yn.cn", | |
"za.com"] | |
def extract_domain(host): | |
""" | |
Domain name extractor. Turns host names into domain names, ported | |
from pwdhash javascript code""" | |
host = re.sub('https?://', '', host) | |
host = re.match('([^/]+)', host).groups()[0] | |
domain = '.'.join(host.split('.')[-2:]) | |
if domain in _domains: | |
domain = '.'.join(host.split('.')[-3:]) | |
return domain | |
_password_prefix = '@@' | |
def generate(password, uri): | |
""" | |
generate the pwdhash password for master password and uri or | |
domain name. | |
""" | |
realm = extract_domain(uri) | |
if password.startswith(_password_prefix): | |
password = password[len(_password_prefix):] | |
password_hash = b64_hmac_md5(password, realm) | |
size = len(password) + len(_password_prefix) | |
nonalphanumeric = len(re.findall(r'\W', password)) != 0 | |
return apply_constraints(password_hash, size, nonalphanumeric) | |
def apply_constraints(phash, size, nonalphanumeric): | |
""" | |
Fiddle with the password a bit after hashing it so that it will | |
get through most website filters. We require one upper and lower | |
case, one digit, and we look at the user's password to determine | |
if there should be at least one alphanumeric or not. | |
""" | |
starting_size = size - 4 | |
result = phash[:starting_size] | |
extras = itertools.chain((ord(ch) for ch in phash[starting_size:]), | |
itertools.repeat(0)) | |
extra_chars = (chr(ch) for ch in extras) | |
nonword = re.compile(r'\W') | |
def next_between(start, end): | |
interval = ord(end) - ord(start) + 1 | |
offset = next(extras) % interval | |
return chr(ord(start) + offset) | |
for elt, repl in ( | |
(re.compile('[A-Z]'), lambda: next_between('A', 'Z')), | |
(re.compile('[a-z]'), lambda: next_between('a', 'z')), | |
(re.compile('[0-9]'), lambda: next_between('0', '9'))): | |
if len(elt.findall(result)) != 0: | |
result += next(extra_chars) | |
else: | |
result += repl() | |
if len(nonword.findall(result)) != 0 and nonalphanumeric: | |
result += next(extra_chars) | |
else: | |
result += '+' | |
while len(nonword.findall(result)) != 0 and not nonalphanumeric: | |
result = nonword.sub(next_between('A', 'Z'), result, 1) | |
amount = next(extras) % len(result) | |
result = result[amount:] + result[0:amount] | |
return result | |
def console_main(): | |
import getpass, sys, os | |
from os import waitpid | |
from subprocess import Popen, PIPE | |
from optparse import OptionParser | |
def parse_cmd_line(): | |
parser = OptionParser(usage=': %prog [-c] | [domain]') | |
parser.add_option('-c', action='store_true', dest='clipboard_domain') | |
parser.set_defaults(clipboard_domain = False) | |
return parser.parse_args() | |
def copy(text): | |
print(text) | |
try: | |
(options, args) = parse_cmd_line() | |
if args != []: | |
domain = sys.argv[1] | |
elif options.clipboard_domain: | |
domain = Popen(["pbpaste"], stdout=PIPE).communicate()[0] | |
else: | |
domain = input("domain: ").strip() | |
password = getpass.getpass("Password for %s: " % domain) | |
generated = generate(password, domain) | |
copy(generated) | |
except BaseException as e: | |
print(e) | |
print('') | |
pass | |
if __name__ == '__main__': | |
console_main() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment