Using the K-means Method to Identify a User by Keyboard Handwriting

Abstract

This article discusses the problem of protection against unauthorized access to data by identifying users by biometric characteristics - keyboard handwriting. To do this, the authors conducted a series of experiments to obtain a set of samples of keyboard handwriting, which are used as a biometric characteristic to identify the owner. To analyze the data and determine the author of a particular handwriting sample, the authors consider the k -means method. This method is one of the simplest and most effective statistical classification methods when the number of clusters is known in advance. To evaluate the effectiveness of this method for solving this problem, the authors propose using the coefficients of false access and false access denial, which are the main characteristics of biometric authentication systems. The results suggest that there are some limitations when using this method. They are related to the fact that this problem is poorly formalized and depends on many factors that can not be mathematically described. For example, the instability of the keyboard handwriting, which is due to changes in the psycho-physiological state of the user, the ergonomics of the keyboard and others. Given these features,the authors propose to solve the problem using methods based on intelligent data processing, which allow to detect hidden patterns and dependencies in the data flow.