Keystroke Dynamics Based User Authentication System

Abstract

Keystroke dynamics refer to the unique patterns of how an individual types on a keyboard. These patterns can be used for biometric identification or authentication, based on the assumption that every user has a distinct typing style. Keystroke dynamics are part of behavioral biometrics, and feature extraction is a crucial step in analyzing keystroke patterns for identification or authentication purposes. In keystroke dynamics, the goal of feature extraction is to capture the distinctive typing behavior of individuals in a numerical or structured format, which can be used in further stages such as classification, verification, or identification and is stored in a database. The two primary purposes of authentication are, firstly, to accurately identify users who are authorized to access private resources and to deny access to those who are not, and secondly, to provide assurance to users that their resources are safeguarded and protected from unauthorized individuals, who often have malicious intent when attempting to access personal resources. Authentication methods have advanced from ancient knocking rhythms to modern face recognition, fingerprint scanning, and signature verification. These methods typically require a combination of two components to verify a user's identity. In this project, users are recognized by identifying their keystroke dynamics in addition to using passwords, employing machine learning and support vector machine classifiers. The typing patterns of users are analyzed to verify their authenticity using machine learning techniques. The Gaussian Mixture Model (GMM), an unsupervised learning method for pattern recognition, is used. The Manhattan distance, also called the taxicab distance or city block distance, calculates the distance between two real-valued vectors (i.e., the distance between keystrokes in this work). Since individuals type in statistically significant yet different ways, an individual's typing pattern differs greatly from that of others, which increases the accuracy of the authentication system.