Hi community!
I recently got in touch with one of your members and he forwarded me to this forum to get a few questions answered. I will shortly introduce myself: I'm a graduate information scientist currently researching the interesting field of 'keystroke dynamics' as part of my MSc dissertation. Since about the early 90's, researchers have been applying machine learning algos to input timings from keyboards to identify (and authenticate) users. This non-intrusive and easy-to-implement type of computer security has gained some steam in recent years, although the practice is still far from perfect.
As a total newbie in the game of keyboards, I have one important question: How precise is a keyboard as a measuring device to conduct research like this with? Are we talking millisecond-accuracy? Microsecond even? I wrote a simple application in JavaScript, which uses the DOMtimeStamp framework to record key input events at 5-microsecond accuracy. Suppose I test my sample of participants using a popular standard, easy-to-get-accustomed-with keyboard (such as MasterKeys Pro L with MX-brown switches), what kind of accuracy can I expect? Does the polling rate of 1000Hz limit the precision of measurement? Are PS2 connections more applicable for this kind of setting? Keep in mind I've already learned a lot about the specifics of keyboard-timings such as debounce delays (which are uniform in my experiment and should therefore not affect results). All I'm interested in is the extent to which I can measure the differences between 2 keystrokes, so in essence: what is the precision of measurement from a keyboard like this?
As an example, I added a graph from an alpha-tester who performed the experiment described above and typed roughly 950 characters. The distribution shows his hold times (dwell times) for each key he pressed, the mean (red) and median (blue). to what extent are measurements like these affected by the accuracy of keyboard timings?
http://imgur.com/a/LttYMLooking forward to learn from you guys!
