A flexible and scriptable password generator which generates strong passphrases, inspired by XKCD 936:
$ xkcdpass > correct horse battery staple
xkcdpass can be easily installed using pip:
pip install xkcdpass
or manually:
python setup.py install
The latest development version can be found on github: https://github.com/redacted/XKCD-password-generator
Contributions welcome and gratefully appreciated!
Python 2 (version 2.7 or later), or Python 3 (version 3.4 or later). Running module unit tests on Python 2 requires mock to be installed.
xkcdpass can be called with no arguments:
$ xkcdpass > pinball previous deprive militancy bereaved numeric
which returns a single password, using the default dictionary and default settings. Or you can mix whatever arguments you want:
$ xkcdpass --count=5 --acrostic='chaos' --delimiter='|' --min=5 --max=6 --valid-chars='[a-z]' > collar|highly|asset|ovoid|sultan > caper|hangup|addle|oboist|scroll > couple|honcho|abbot|obtain|simple > cutler|hotly|aortae|outset|stool > cradle|helot|axial|ordure|shale
which returns
--count=55 passwords to choose from--acrostic='chaos'the first letters of which spell 'chaos'--delimiter='|'joined using '|'--min=5 --max=6with words between 5 and 6 characters long--valid-chars='[a-z]'using only lower-case letters (via regex).
A concise overview of the available xkcdpass options can be accessed via:
xkcdpass --help
Usage: xkcdpass [options]
Options:
-h, --help
show this help message and exit
-w WORDFILE, --wordfile=WORDFILE
Specify that the file WORDFILE contains the list of
valid words from which to generate passphrases. Multiple
wordfiles can be provided, separated by commas.
Provided wordfiles: eff-long (default), eff-short,
eff-special, legacy, spa-mich (Spanish), fin-kotus (Finnish)
ita-wiki (Italian), ger-anlx (German), nor-nb (Norwegian),
fr-freelang (French), pt-ipublicis / pt-l33t-ipublicis (Portuguese)
--min=MIN_LENGTH
Minimum length of words to make password
--max=MAX_LENGTH
Maximum length of words to make password
-n NUMWORDS, --numwords=NUMWORDS
Number of words to make password
-i, --interactive
Interactively select a password
-v VALID_CHARS, --valid-chars=VALID_CHARS
Valid chars, using regexp style (e.g. '[a-z]')
-V, --verbose
Report various metrics for given options, including word list entropy
-a ACROSTIC, --acrostic=ACROSTIC
Acrostic to constrain word choices
-c COUNT, --count=COUNT
number of passwords to generate
-d DELIM, --delimiter=DELIM
separator character between words
-R, --random-delimiters
use randomised delimiters
-D DELIMITERS, --valid-delimiters=DELIMETERS
delimeters to choose from, used with -
-s SEP, --separator SEP
Separate generated passphrases with SEP.
-C CASE, --case CASE
Choose the method for setting the case of each word in
the passphrase. Choices: ['alternating', 'upper',
'lower', 'random', 'capitalize'] (default: 'lower').
--allow-weak-rng
Allow fallback to weak RNG if the system does not
support cryptographically secure RNG. Only use this if
you know what you are doing.
Several word lists are provided with the package. The default, eff-long, was specifically designed by the EFF for passphrase generation and is licensed under CC BY 3.0. As it was originally intended for use with Diceware ensure that the number of words in your passphrase is at least six when using it. Two shorter variants of that list, eff-short and eff-special, are also included. Please refer to the EFF documentation linked above for more information.
The original word list from xkcdpass versions earlier than 1.10.0 is also provided as a convenience, and is available under legacy. This word list is derived mechanically from 12Dicts by Alan Beale. It is the understanding of the author of xkcdpass that purely mechanical transformation does not imbue copyright in the resulting work. The documentation for the 12Dicts project at
http://wordlist.aspell.net/12dicts/ contains the following dedication:
The 12dicts lists were compiled by Alan Beale. I explicitly release them to the public domain, but request acknowledgment of their use.
Note that the generator can be used with any word file of the correct format: a file containing one 'word' per line.
- Spanish: a modifed version of archive.umich.edu in the /linguistics directory. It includes ~80k words. Less than 5 char. and latin-like words were deleted using regex. This list is public domain, see here.
- Finnish: a modified version of the Institute for the Languages of Finland XML word list. Profanities and expressions containing spaces were removed using regex. The resulting list contains ~93k words. The list is published under GNU LGPL, EUPL 1.1 and CC-BY 3.0 licenses.
- Italian: generated from dumps of the Italian-language Wikipedia, which is released under the Creative Commons Attribution-Share-Alike 3.0 licence.
- German (ger-anlx): based on this GPL v3 list. Single and double character words have been removed.
- German (eff_large_de_sample.wordlist): based on this public domain dictionary. Converted to UTF-8 and randomly sampled to reduce file size.
- Norwegian: a modified version of Norsk Ordbank in Norwegian Bokmål 2005, 2018-06-28 update, which is released under the CC-BY 4.0 license. Regex has been used to alter the list for cleanup and removal of words with impractical characters. The resulting list contains ~137k words.
- French: One cleaned version of this list (public domain), and one filtered to remove potentially offensive words.
- Portuguese: Converted variant of the LibreOffice / Firefox poturguese dictionary (from this link. GPL and BSD licenced.
Additional language word lists are always welcome!
The built-in functionality of xkcdpass can be extended by importing the module into python scripts. An example of this usage is provided in example_import.py, which randomly capitalises the letters in a generated password. example_json.py demonstrates integration of xkcdpass into a Django project, generating password suggestions as JSON to be consumed by a Javascript front-end.
A simple use of import:
from xkcdpass import xkcd_password as xp # create a wordlist from the default wordfile # use words between 5 and 8 letters long wordfile = xp.locate_wordfile() mywords = xp.generate_wordlist(wordfile=wordfile, min_length=5, max_length=8) # create a password with the acrostic "face" print(xp.generate_xkcdpassword(mywords, acrostic="face"))
When used as an imported module, generate_wordlist() takes the following args (defaults shown):
wordfile=None, min_length=5, max_length=9, valid_chars='.'
While generate_xkcdpassword() takes:
wordlist, numwords=6, interactive=False, acrostic=False, delimiter=" "
xkcdpass uses crytographically strong random number generators where possible (provided by random.SystemRandom() on most modern operating systems). From version 1.7.0 falling back to an insecure RNG must be explicitly enabled, either by using a new command line variable before running the script:
xkcdpass --allow-weak-rng
or setting the appropriate environment variable:
export XKCDPASS_ALLOW_WEAKRNG=1
- 1.19.3 Restore a randomly sampled version of eff_large_de wordlist
- 1.19.2 Reduction in install size
- 1.19.1 Improvements to help text, handle rare case where arguments lead to empty wordlist
- 1.19.0 Initial support for multiple wordfiles
- 1.18.2 fixes for README
- 1.18.0 Added randomised delimiters
- 1.17.6 Bugfixes
- 1.17.5 Bugfixes
- 1.17.4 Improvements to French dictionary
- 1.17.3 Updated license and supported versions
- 1.17.2 Compatibility fix for 2.x/3.x
- 1.17.1 Fix issue with README and unicode encoding
- 1.17.0 Add French, Norwegian, and Portuguese dictionaries. Bugfixes and improvements to tests (WIP).
- 1.16.5 Adds title case option for --case
- 1.16.4 Improve unit tests, fixes broken test on python 2
- 1.16.3 Correct links for German worldist, updated docs to include the list
- 1.16.2 Fix exception on UTF8 open with python 2.x
- 1.16.1 Fix encoding issue on Windows
This is free software: you may copy, modify, and/or distribute this work under the terms of the BSD 3-Clause license.
See the file LICENSE.BSD for details.