User-representative feature selection for keystroke dynamics

摘要: Continuous user authentication with keystroke dynamics uses characters sequences as features. Since users can type in any order, it is imperative to find character (n-graphs) that are representative of typing behavior. The contemporary feature selection approaches do not guarantee selecting frequently-typed features which may cause less accurate statistical user-representation. Furthermore, the selected inherently reflect We propose four statistical-based techniques mitigate limitations existing approaches. first technique selects most frequently occurring other three consider different behaviors by selecting: n-graphs typed quickly; consistent time; and have large time variance among users. use Gunetti's dataset k-means clustering algorithm for our experiments. results show proposed techniques, most-frequent effectively user-representative further substantiate comparing (popular Italian words, common n-graphs, least frequent n-graphs). performs better than after a certain number n-graphs.