Dynamic-signature-based user authentication using a fuzzy classifier

Dynamic signature verification is one of the most fast, intuitive, and cost effective tools for user authentication. Dynamic signature recognition uses multiple characteristics in the analysis of an individual's handwriting. Dynamic characteristics include the velocity, acceleration, timing, pressure, and direction of the signature strokes, all analyzed in the x, y, and z directions. In this paper, the constant term and the first seven harmonics of the Fourier series expansion of the signature were used as features. The authentication systems development includes the following stages: preprocessing, feature selection, classification. Binary metaheuristic algorithms and deterministic algorithms are used to select attributes. The classification was carried out using a fuzzy classifier. The fuzzy classifiers parameters were tuned using continuous metaheuristic algorithms. The efficiency of the authentication system was verified on the author's database. The database contains 280 original variants of the signature of one author and 1281 variants of counterfeit signatures of seven authors. To assess the statistical significance of differences in the accuracy and error rates of the fuzzy classifiers formed by metaheuristic algorithms, the Mann-Whitney (-Wilcoxon) U-test to compare medians and the Kruskal-Wallis test were used.