A primitive write-up of the Safelock system is available here - complete with ROC curve and some performance metrics.
I've included a little more detail about the four keystroke attributes we measured as well as a surface-level explanation of the algorithm
Nonetheless, my primary concern would be that it would lock people out of their computers/applications when they have had a little much to drink.
We had a >95% True Positive rate (with a >99% True Negative. I can dig up the ROC curve if you really care...). Basically, the idea is to find and measure typing attributes that are keyboard/mood/alchohol level-agnostic.
We're still working on getting funding for testing the algorithm after a few drinks, though.
Slashdot Mirror
A primitive write-up of the Safelock system is available here - complete with ROC curve and some performance metrics. I've included a little more detail about the four keystroke attributes we measured as well as a surface-level explanation of the algorithm
Nonetheless, my primary concern would be that it would lock people out of their computers/applications when they have had a little much to drink.
We had a >95% True Positive rate (with a >99% True Negative. I can dig up the ROC curve if you really care...). Basically, the idea is to find and measure typing attributes that are keyboard/mood/alchohol level-agnostic. We're still working on getting funding for testing the algorithm after a few drinks, though.
I like the idea, but if they remove the *click* physical feedback, someone's going to die.
The keyboard still has the rubber domes, just like all the others.