Analysis of Adaptive behavioral Authentication Model

This thesis is devoted to the synthesis and analysis of keystroke dynamics authentication. The work in this thesis examines the keystroke dynamics as a behavioral authentication. Keystroke dynamics is one of the famous behavioral measurements that rely on utilizing the typing rhythm of each individual. It is used to strengthen access control mechanism by increasing authentication power efficiently and cheaply since no additional hardware is needed. The study focuses on enhancing security level and strengthening access control using artificial neural networking model based on multilayer perceptron classifier. Multilayer perceptron classifier model which uses back propagation algorithm for weights adaptation during learning and testing phases is proposed. The proposed model is trained and tested using real dataset samples: which were collected by the proposed designed virtual keyboard. It was installed on specific Android based touch screen device (Sony Xperia tablet Z). The collected dataset records consist of timing and non-timing data for typing rhythm of each user. This thesis presents a unique approach for combining timing and non-timing features together.