Keystroke identification with a genetic fuzzy classifier

This paper proposes the use of fuzzy if-then rules for Keystroke identification. The proposed methodology modifies Ishibuchi's genetic fuzzy classifier to handle high dimensional problems such as keystroke identification. High dimensional property of a problem increases the number of rules with low fitness. For decreasing them, rule initialization and coding are modified. Furthermore a new heuristic method is developed for improving the population quality while running GA. Experimental result demonstrates that we can achieve better running time, interpretability and accuracy with these modifications.