How do you type? Do you type in the same way, every time you type your email address, your password? Chances are, you do. Chances are, you use a pattern, which is repeatable and always the same (for the most part).
It’s like your signature: For my own, I first write all the lower case letters and then put the dots on the i and the line on the t. It’s how I write. It’s how I identify myself.
And now, there is a new proposed software that will allow security systems to verify you by utilizing that pattern to identify you: Keystroke Dynamics.
Several scientific papers have been written about a new algorithm that proposes to combine already existing software to make a new type of security feature: a real “signature” that identifies the nature of HOW we type, not just the outcome of WHAT we have typed. Our “handwriting” is analysed.
A program, which learns how you type, where you hesitate, which patterns are uniquely yours. It is, essentially, a biometric like a retina scan or a finger print – or your real-life signature.
As I am typing this, I am acutely aware that I always tend to hesitate before I type the y. This new security feature would have the user type a text, (the longer the better) and learns how you type. You would then have to re-type your proposed password (or pass-text) several times so the algorithm learns your patterns: Where you hesitate, which letters and words you type fastest, how your fingers work across the keyboard.
Then, it would add another layer of security against bots and hackers.
This method can be implemented with little to no costs, as the software it runs on already exists in computers. It would just mean adding another layer into the signup and verification process for new users.
This is a very new field of research but definitely one that does need further testing and resources before it can be implemented on a larger scale. The scientists who proposed this new method are J. Visumathi & P. Jesu Jayarin of and they first published their research in the Journal of Applied Security Research in July.
In their paper, they state that “The proposed approach is easy to understand why keystroke biometrics using Mahalanobis distance and Manhattan distance outperformed other algorithms including some of the more advanced machine learning techniques. The keystroke dynamics features consist of both dwell and latency timings to have large variations in individual components. Each of the two metrics, when used alone, has its advantages and limitations.”
The detailed proposition can be found on this link and definitely makes for a hugely interesting read.