Gait analysis as a prominent behavioral biometric modality for customer authentication purposes

Abstract

As technology continues to advance, people are increasingly interacting with a variety of systems, ranging from mobile phones and wearables to payment terminals and vehicles. Although many of these systems still rely on traditional forms of authentication, such as PINs and passwords, these methods are not without limitations due to human errors and biases. As a result, there is a growing interest in the use of Behavioral Biometrics (BB) as a more seamless and passive authentication method. BB is typically used in combination with multifactor authentication (MFA). In this thesis, the focus is on gait analysis using accelerometer and gyroscope data as a biometric modality. The study aims to investigate the effectiveness of classical machine learning techniques in analyzing gait data for authentication purposes, as well as the potential of gait-based BB to enhance customer authentication through MFA. The study is divided two parts: a theoretical exploration of BB, and an empirical analysis of gait data. A combination of qualitative and quantitative research methods was employed to investigate the effectiveness of various machine learning models and their applicability in the context of BB.