Voice Identification of Spanish-Speakers Using a Convolution Neural Network in the Audio Interface of a Computer Attack Analysis Tool

This work is dedicated to the elaboration of a machine learning model for identifying users by voice. The voice identification module complements the previously developed prototype of a software solution for professional monitoring and response to information security incidents for visually impaired people. A brief characteristic of existing assistive technologies is presented. The article highlights the conditions and procedure of the experiment that was carried out in order to implement voice authorization in every day information security routine at one of the Universities. The main aim of the elaborated voice recognition module is, on the one hand simplify usage of information security software, on other is to implement the principal of employee inclusion in cybersecurity. Speech recordings of nine Spanish-speaking university employees were used as a training sample. Combinations of features were studied, according to which the convolution neural network has classified speakers with 90% accuracy.