IT WAS the best of times, it was the worst of times, then you forgot your password and it totally went to pot. Thankfully, a way of generating secure but memorable passwords based on the work of Charles Dickens 鈥� or indeed any large text 鈥� will make things far better.
People are terrible at choosing passwords 鈥� 鈥�123456鈥� and 鈥減assword鈥� are among the most used 鈥� so ideally we should randomly generate them. Something like 鈥淭H9H*kt7鈥� is very secure because the mix of upper case, lower case and special characters makes it just one of 72 quadrillion possible 8-character passwords, and hackers often start by guessing words.
But it鈥檚 hard to remember that garbled mess. That鈥檚 why John Clements of California Polytechnic State University in San Luis Obispo has turned to Dickens. He used a tool from probability theory called a Markov model. The model looked at pairs of characters 鈥� letters and punctuation, not fictional people 鈥� in the book A Tale of Two Cities, and saw how often other characters appeared after them. Taking 鈥渃a鈥� as an example, 鈥渃ar鈥� is more likely than 鈥渃af鈥�.
Advertisement
To build passwords, the model takes two characters and generates a third. The second and third characters are then put back into the model to give a fourth, and so on until you have a password that is long enough to be secure ().
The passwords contain words that seem memorable and pronounceable, even if they aren鈥檛 real English, such as 鈥淭herying hant abree,鈥�. Starting with three characters or more instead of pairs results in more familiar words, at the cost of length, giving things like 鈥淭he shing it nother to delve w鈥�.
These passwords are just as secure as 鈥淭H9H*kt7鈥�, and hopefully easier to remember. Clements plans to explore using a person鈥檚 email history to generate passwords, so they will be based on personal idioms, rather than those of Dickens.
This article appeared in print under the headline 鈥淟et Charles Dickens sort out your passwords鈥�
