Using molecular algorithm in keystroke dynamics

Previous studies on keystroke dynamics confirmed that the use of biometric technologies provides a significantly high level of reliability especially in situations where security is paramount. Most of these studies use genetic algorithms and k-nearest neighbor as their model of computation. In this paper, we explore an algorithmic solution based on an alternative model patterned after the process of chemical reactions of molecules. We invited several respondents to participate in our testing. Multiple keystroke samples were taken from each respondent and their keystroke values were stored in our databank. Samples collected in this paper were modeled using periodic tables and molecular bonding of keystrokes as data representations. In performing the computation, a molecular algorithm was introduced. A molecular bonding was constructed to stimulate chemical reactions such as synthesis, decomposition and displacement. These molecular reactions enhance the pattern matching process of a keystroke against the collection of samples. Experiments show that the algorithm demonstrates 92% correctness in authenticating a keystroke pattern, while it can fairly recognize impostors during the identification process.