randword: Generating memorable random passwords

This started when I decided to learn python by rewriting one of my old perl scripts in python.  randstring is a script to generate a random string of characters.  I use it sometimes to generate passwords, but password strings or random characters usually can't be remembered, at least not easily.  Passwords like that can be useful at times.  You need to store them in an encrypted password safe.

I have another script that generates more memorable passwords.  Some people I know, have found it useful.  There are always some passwords you need to be memorable.  For instance your login password and the password to your password safe.  randword generates a bunch of words from a dictionary.  XKCD style passwords, if you like.  In the process of examining it, I rewrote it in both perl and python, fixed some bugs and added some features.

In general a bunch of words can be much easier to remember and can be just as difficult or far more difficult  to crack.  I like to generate a bunch and choose a few at random.  4 or 5 or more words is OK.  Hint: misspellings are good but not if you can't remember what you did.  Passwords on websites are a bit mad at the moment with complicated rules, like: "there's an illegal character" or "you must have an upper-case letter and a number", or "that password is too short", or "too long" etc. 

New features of randword:

• There's a couple of new options about output format, like camel case.   
• randword can use any dictionary or word frequency lists as long as they have a fairly simple format - ie at least a word and an optional number at the start of each line. 
• randword can also take a bunch of text and create dictionaries of words that it can use to generate random passwords.  

For word lists, I have used various texts, for instance Jane Austen's complete works, Shakespeare, Mark Twain, Chaucer.  There are many works that can be easily got from Project Gutenburg among other places on the net.  Also word lists and text that can be found at COCA or Lancaster University  etc.Since I only want ascii because I can't type non-ascii characters easily, I used unidecode (python or original perl version) to turn them into ascii.  Python unidecode comes with a command line script.  I wrote a very simple perl script to detect non-ascii characters, (not included) although working out what encoding a page is in is a kind of major headache and you need to know the encoding before unidecode will work, grrrr. 

The links below include word lists from Chaucer, Shakespeare, Mark Twain, and the linux word dictionary.

This is my original blog post on the scripts with all the links to the scripts and associated stuff.

Links:
randstring.pl randstring.py
randword.pl randword.py
some word lists
tarred and zipped archive of scripts and wordlists

ngraph: another random text generator.

In my series of scripts to generate random text for fun and for helping create secure passwords this is my latest.  This new set of scripts generate text based on frequency weighted random choice of ngraphs.  I use it sometimes to generate word-like things  to create passwords.  There may be other uses, for instance creating random text with English characteristics.

ngraph:  Generating text from frequency weighted letter combinations

An ngraph is a group of n consecutive letters occurring in a language.  (This is my definition, there may be another word for it but I couldn't find it.) A set of ngraph frequencies is a set of the number of times each ngraph is used in a group of texts.  So for instance for n=1 we have the frequencies of the letters.  For n=2 we have the set of frequencies of the digraphs, for n=3, the trigraphs etc.  Because it is easy to do on a computer, I include some punctuation (space return - ' , . ; ! ? &).  I use these files among others:
The complete works of Jane Austen and the complete works of William Shakespeare.
Any texts would do.  Gutenburg texts have a few oddities that the script is designed to work with.

There are two main scripts and a subsidiary script.  (ngraph.pl, dbfill.sh and ngraph-db.pl)

The first script is "ngraph.pl": this script has two separate functions.
Firstly it reads a series of files with presumably text in a human language and generates a set of ngraphs for an "n" you specify.
Then ngraph.pl will generate a random set of text based on the ngraphs and their frequency.

Alternatively it can just output the set of ngraph frequencies as text or sql.  The reason for this is that reading the files and creating ngraph frequency tables is a resource intensive process so I decided to create a database of ngraphs and to generate text from that database.  I found a database with n = 1 to 5 to be most useful and above 5 the amount of data gets massive and more actual words are generated. 

The second script "dbfill.sh" is a subsidiary script.  It creates the database and populates it with ngraphs using the first script.

The third script is "ngraph-db.pl"
This uses the database and generates text based on the ngraphs in the database.  Because it has access to a database with ngraphs of say n = 1 to 5 it can generate text from random sized ngraphs as well as a single ngraph.

The generated text can include words but it mostly has word-like things that are a bit memorable but not actual words.  I never use the generated text directly to create passwords but pick and choose bits and let parts of the text inspire a password.

The scripts are available under the GPL here.

Here is a sample output:

$ngraph-db.pl -W -c 1000 -2 -g 1-4