realistic password strength estimation


zxcvbn is a password strength estimator inspired by password crackers. Through pattern matching and conservative entropy calculations, it recognizes and weighs 30k common passwords, common names and surnames according to US census data, popular English words from Wikipedia and US television and movies, and other common patterns like dates, repeats (aaa), sequences (abcd), keyboard patterns (qwertyuiop), and l33t speak.

Consider using zxcvbn as an algorithmic alternative to password composition policy — it is more secure, flexible, and usable when sites require a minimal complexity score in place of annoying rules like "passwords must contain three of {lower, upper, numbers, symbols}".

  • More secure: policies often fail both ways, allowing weak passwords (P@ssword1) and disallowing strong passwords.
  • More flexible: zxcvbn allows many password styles to flourish so long as it detects sufficient complexity — passphrases are rated highly given enough uncommon words, keyboard patterns are ranked based on length and number of turns, and capitalization adds more complexity when it's unpredictaBle.
  • More usable: Use zxcvbn to build simple, rule-free interfaces that give instant feedback. In addition to scoring, zxcvbn includes minimal, targeted verbal feedback that can help guide users towards less guessable passwords.

At Dropbox we use zxcvbn on our signup page and change/reset password flows. zxcvbn is designed for node and the browser, but we use our python port inside the Dropbox desktop client, Objective C port in our iOS app, and Java port (not yet open sourced) on Android.

Release notes

For more motivation, see:



zxcvbn detects and supports CommonJS (node, browserify) and AMD (RequireJS). In the absence of those, it adds a single function zxcvbn() to the global namespace.

Install node and bower if you haven't already.

Get zxcvbn:

cd /path/to/project/root
bower install zxcvbn

Add this script to your index.html:

<script type="text/javascript" src="bower_components/zxcvbn/dist/zxcvbn.js">

To make sure it loaded properly, open in a browser and type zxcvbn('Tr0ub4dour&3') into the console.

To pull in updates and bug fixes:

bower update zxcvbn

zxcvbn works identically on the server.

$ npm install zxcvbn
$ node
> var zxcvbn = require('zxcvbn');
> zxcvbn('Tr0ub4dour&3');

Meteor (via Atmosphere)

meteor add codetheweb:zxcvbn

Add zxcvbn.js to your project (using bower, npm or direct download) and import as usual:

requirejs(["relpath/to/zxcvbn"], function (zxcvbn) {

If you're using npm and have require('zxcvbn') somewhere in your code, browserify and webpack should just work.

$ npm install zxcvbn
echo "console.log(require('zxcvbn'))" > mymodule.js
$ browserify mymodule.js > browserify_bundle.js
$ webpack mymodule.js webpack_bundle.js

But we recommend against bundling zxcvbn via tools like browserify and webpack, for three reasons:

  • Minified and gzipped, zxcvbn is still several hundred kilobytes. (Significantly grows bundle size.)
  • Most sites will only need zxcvbn on a few pages (registration, password reset).
  • Most sites won't need zxcvbn() immediately upon page load; since zxcvbn() is typically called in response to user events like filling in a password, there's ample time to fetch zxcvbn.js after initial html/css/js loads and renders.

See the performance section below for tips on loading zxcvbn stand-alone.

Tangentially, if you want to build your own standalone, consider tweaking the browserify pipeline used to generate dist/zxcvbn.js:

$ browserify --debug --standalone zxcvbn \
    -t coffeeify --extension='.coffee' \
    -t uglifyify \
    src/main.coffee | exorcist dist/zxcvbn.js.map >| dist/zxcvbn.js
  • --debug adds an inline source map to the bundle. exorcist pulls it out into dist/zxcvbn.js.map.
  • --standalone zxcvbn exports a global zxcvbn when CommonJS/AMD isn't detected.
  • -t coffeeify --extension='.coffee' compiles .coffee to .js before bundling. This is convenient as it allows .js modules to import from .coffee modules and vice-versa. Instead of this transform, one could also compile everything to .js first (npm run prepublish) and point browserify to lib instead of src.
  • -t uglifyify minifies the bundle through UglifyJS, maintaining proper source mapping.

Download zxcvbn.js.

Add to your .html:

<script type="text/javascript" src="path/to/zxcvbn.js">


try zxcvbn interactively to see these docs in action.

zxcvbn(password, user_inputs=[])

zxcvbn() takes one required argument, a password, and returns a result object with several properties:

result.guesses            # estimated guesses needed to crack password 
result.guesses_log10      # order of magnitude of result.guesses 
result.crack_time_seconds # dictionary of back-of-the-envelope crack time 
                          # estimations, in seconds, based on a few scenarios: 
  # online attack on a service that ratelimits password auth attempts. 
  # online attack on a service that doesn't ratelimit, 
  # or where an attacker has outsmarted ratelimiting. 
  # offline attack. assumes multiple attackers, 
  # proper user-unique salting, and a slow hash function 
  # w/ moderate work factor, such as bcrypt, scrypt, PBKDF2. 
  # offline attack with user-unique salting but a fast hash 
  # function like SHA-1, SHA-256 or MD5. A wide range of 
  # reasonable numbers anywhere from one billion - one trillion 
  # guesses per second, depending on number of cores and machines. 
  # ballparking at 10B/sec. 
result.crack_time_display # same keys as result.crack_time_seconds, 
                          # with friendlier display string values: 
                          # "less than a second", "3 hours", "centuries", etc. 
result.score      # Integer from 0-4 (useful for implementing a strength bar) 
  0 # too guessable: risky password. (guesses < 10^3) 
  1 # very guessable: protection from throttled online attacks. (guesses < 10^6) 
  2 # somewhat guessable: protection from unthrottled online attacks. (guesses < 10^8) 
  3 # safely unguessable: moderate protection from offline slow-hash scenario. (guesses < 10^10) 
  4 # very unguessable: strong protection from offline slow-hash scenario. (guesses >= 10^10) 
result.feedback   # verbal feedback to help choose better passwords. set when score <= 2. 
  result.feedback.warning     # explains what's wrong, eg. 'this is a top-10 common password'. 
                              # not always set -- sometimes an empty string 
  result.feedback.suggestions # a possibly-empty list of suggestions to help choose a less 
                              # guessable password. eg. 'Add another word or two' 
result.sequence   # the list of patterns that zxcvbn based the 
                  # guess calculation on. 
result.calc_time  # how long it took zxcvbn to calculate an answer, 
                  # in milliseconds. 

The optional user_inputs argument is an array of strings that zxcvbn will treat as an extra dictionary. This can be whatever list of strings you like, but is meant for user inputs from other fields of the form, like name and email. That way a password that includes a user's personal information can be heavily penalized. This list is also good for site-specific vocabulary — Acme Brick Co. might want to include ['acme', 'brick', 'acmebrick', etc].


zxcvbn operates below human perception of delay for most input: ~5-20ms for ~25 char passwords on modern browsers/CPUs, ~100ms for passwords around 100 characters. To bound runtime latency for really long passwords, consider sending zxcvbn() only the first 100 characters or so of user input.

zxcvbn.js bundled and minified is about 390kb gzipped or 800kb uncompressed, most of which is dictionaries. Consider these tips if you're noticing page load latency on your site.

Then try one of these alternatives:

  1. Put your <script src="zxcvbn.js"> tag at the end of your html, just before the closing </body> tag. This insures your page loads and renders before the browser fetches and loads zxcvbn.js. The downside with this approach is zxcvbn() becomes available later than had it been included in <head> — not an issue on most signup pages where users are filling out other fields first.

  2. If you're using requirejs, try loading zxcvbn.js separately from your main bundle. Something to watch out for: if zxcvbn.js is required inside a keyboard handler waiting for user input, the entire script might be loaded only after the user presses their first key, creating nasty latency. Avoid this by calling your handler once upon page load, independent of user input, such that the requirejs() call runs earlier.

  3. Use the HTML5 async script attribute. Downside: doesn't work in IE7-9 or Opera Mini.

  4. Include an inline <script> in <head> that asynchronously loads zxcvbn.js in the background. Despite the extra code I prefer this over (3) because it works in older browsers.

// cross-browser asynchronous script loading for zxcvbn. 
// adapted from http://friendlybit.com/js/lazy-loading-asyncronous-javascript/ 
(function() {
  var ZXCVBN_SRC = 'path/to/zxcvbn.js';
  var async_load = function() {
    var first, s;
    s = document.createElement('script');
    s.src = ZXCVBN_SRC;
    s.type = 'text/javascript';
    s.async = true;
    first = document.getElementsByTagName('script')[0];
    return first.parentNode.insertBefore(s, first);
  if (window.attachEvent != null) {
    window.attachEvent('onload', async_load);
  } else {
    window.addEventListener('load', async_load, false);


Bug reports and pull requests welcome!

git clone https://github.com/dropbox/zxcvbn.git

zxcvbn is built with CoffeeScript, browserify, and uglify-js. CoffeeScript source lives in src, which gets compiled, bundled and minified into dist/zxcvbn.js.

npm run build    # builds dist/zxcvbn.js 
npm run watch    # same, but quickly rebuilds as changes are made in src. 

For debugging, both build and watch output an external source map dist/zxcvbn.js.map that points back to the original CoffeeScript code.

Two source files, adjacency_graphs.coffee and frequency_lists.coffee, are generated by python scripts in data-scripts that read raw data from the data directory.

For node developers, in addition to dist, the zxcvbn npm module includes a lib directory (hidden from git) that includes one compiled .js and .js.map file for every .coffee in src. See prepublish in package.json to learn more.


Dropbox for supporting open source!

Leah Culver and Ryan Pearl for porting zxcvbn to Objective C and python.

Mark Burnett for releasing his 10M password corpus and for his 2005 book, Perfect Passwords: Selection, Protection, Authentication.

Wiktionary contributors for building a frequency list of English as used in television and movies.

Researchers at Concordia University for studying password estimation rigorously and recommending zxcvbn.

And xkcd for the inspiration :+1::horse::battery::heart: