Wouldn't this be fooled by recording of mouse movements. I can imagine the following data points without specialist mouse hardware.
* the rate of acceleration and deceleration as you move from the original cursor point to the target.
* the angle of the arc of movement between the two points.
* the delay between movement ceasing and clicking
* double click profile (time between each click and how still you can keep the mouse)
OK, so plug those into some algorithm and give a score as to how likely it is the same user.
Now do all that again and imagine some malware software is recording your mouse movement profile (could even be embedded in a freebee mouse). A vnc style piece of software could after not too much time now allow you to perform an action but instead emulate the recorded profile in those actions.
Not as trivial as a rainbow table, but if these techniques take off you can bet such tools will become available.