Using current methods it would be difficult to uniquely identify an individual user out of all based solely on their typing pattern (because entropy, they aren't unique enough). But you could fairly reliably use it as a 2nd factor to verify someone as themselves vs. a malicious user