Secure random passwords in javascript
HTML Python JavaScript
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
js
markov_corpus
txt
LICENSE
README.md
blank
build.py
cache.manifest
favicon.ico
fusionbox.gif
index.html
markov.py
mouseware.py
words.py

README.md

Mouseware uses a cryptographically secure random number generator based on your mouse movements to generate secure, memorable passwords. Passwords are generated entirely in the browser, no data is ever sent over the network. The generated passphrases are similar to those generated by Diceware or popularized by xkcd, with an emphasis on easy memorization.

We use SHA256 to process the location and timestamp of every mousemove event. Random words are then selected from noun, adjective and verb word lists to construct a random sentence. Shannon entropy is reported to estimate password strength. To avoid a time-consuming mouse moving session on every page load, we use localStorage to store a seed and initialize our internal entropy buffer with it on the next load.

To increase entropy when desired, options to replace a randomly chosen letter with a number or symbol are given. Entropy calculation takes these into account, and increases the reported entropy by the logarithm of the number of letters considered for replacement.

Mouseware also provides an alternative markov chain based passphrase generator. This generator uses a markov chain to generate passphrases that have a similar structure to normal language, but which doesn't necessarily include actual English words. These passphrases should be more random for a given length than the standard mouseware passphrases, but still relatively easy to remember and to type. This generator also provides an estimated entropy based on the entropy of the series of nodes visited in the markov chain. Note however, that since the markov chain generation process makes a series of decisions which is path dependent, and where the probabilities are not uniform, there is not a simple mapping from this estimated entropy to the average guesswork required to guess a passphrase. As a result, the markov chain approach is likely to generate better passwords for a given length on average. However, if you need very strong guarantees of the strengths of all passwords, the default algorithm is a better choice.