Consider: each person's typing style is unique in overall speed, speed between letter pairs, use of the backspace key, and so on. Profile these typing characteristics for ten people and you, too, can tell who's on the keyboard just by looking at these statistics.
This falls under the category of what I call "behavioral metrics" for identification of end users. A similar effort is underway by law enforcement authorities to identify people by the way they walk (no kidding: it's called "gait analysis" and it's based on the idea that people have unique walks).
Naturally, there are also the usual biometric methods for identification: voiceprint, fingerprint, retinal scans and so on. And there's the lowly password, which turns out to be extremely insecure in real world use due to the fact that most people use passwords that are simultaneously easy to remember and easy to guess.
But typing behavior analysis offers something much better: an innate password, of sorts, that requires no memorization and is virtually impossible to forge. Consider: could you sit down at a keyboard and type like another person? Exactly like another person? I doubt it. Unless your motor skills match theirs almost perfectly, there are going to be rather obvious differences between your typing patterns and theirs. Furthermore, it seems virtually impossible to learn how another person types simply by watching them type, so unlike passwords which are rendered useless as soon as somebody watches you type one in, typing pattern analysis remains secure even when conducted in full view! One more thing: it can be done in software only, so the costs of this technology are by definition dramatically lower than anything requiring hardware (fingerprint scanning, retinal scanning, or other biometric security systems). This results in typing behavior security having unique advantages that simply aren't present in other approaches to computer security and user identification. The only drawback (and yes, there is one) is that identification is nowhere near instant. It takes time (up to 60 seconds, perhaps) for the software to gather enough typing data to make a decision on user authentication. And, of course, there's the problem of typing when you've just come in from a Winter blizzard. Frozen hands don't work as well as warm, fluid ones. Will this software lock you out of your own computer until you warm up with a mug of hot chocolate? Might make a nice excuse at work, but this would hardly win customers at the corporate level. We'll have to watch this one and see how it turns out in the real world.
Researchers at the Technion-Israel Institute of Technology have developed software that is able to identify computer users -- with high accuracy -- by their individual, distinct typing styles. The system prototype was developed in the Data Mining Lab of the Technion's Computer Science Faculty by students Ido Yariv and Mordechai Nisenson, under the supervision of Technion Professors Ran El-Yaniv and Ron Meir. He goes on to explain that time differential patterns between consecutive keystrokes can uniquely determine a user. Many processes and factors make up this sequence, including language fluency, mental attributes, fine motor skills, hand size and finger length.