IJARCCE International Journal of Advanced Research in Computer and Communication Engineering Vol. 10, Issue 9, September 2021 DOI: 10.17148/IJARCCE.2021.10903 @ IJARCCE This work is licensed under a Creative Commons Attribution 4.0 International License 18 ISSN (O) 2278-1021, ISSN (P) 2319-5940 An Investigation on the Impact of Age Group and Gender on the Authentication Performance of Keystroke Dynamics Ademola O. Adesina 1 , Olasupo Oyebola 2 Lecturer, Computer Science Department, Olabisi Onabanjo University, Ago-Iwoye, Nigeria 1 Graduate Student, Computer Science Department, National Open University, Ibadan, Nigeria 2 Abstract: Keystroke dynamics is a biometric that has been explored as a means of making user authentication more secure. However, studies have indicated that the performance of such a system might be influenced by the demography of the user population. The purpose of this study is to investigate the relationship between the age and gender of the users of a keystroke dynamics-based mobile phone user authentication and the performance of the scheme. Using a mobile keystroke dynamics dataset containing the age and gender information of the participants, an anomaly detector algorithm was used to test whether an impostor user would have been recognised or not. A False Acceptance Rate (FAR) is calculated for the genuine user and impostors' combination. A Two-Way Analysis of Variance (ANOVA) was used to test the hypotheses whether there are significance differences and interaction between the FARs obtained with respect to the age group and gender categories. The result suggests that the age and gender of the users of a keystroke dynamics user authentication system on a mobile phone is not expected to have significant impact on the performance of such a system. Unlike previous studies that were based on keystroke dynamics data from desktop computer users, this investigation focused on keystroke dynamics for mobile phones. The results obtained in this paper has further improved our understanding that demographic bias relating to age and gender may be eliminated from the concerns that may arise from the use of a keystroke dynamics user authentication on a mobile phone. Keywords: Biometrics, User Authentication, Keystroke Dynamics, Classification, Machine Learning. I. INTRODUCTION User authentication involves a person proving a claimed identity on a device such as a mobile phone or on a computer. Traditional authentication can be done by the person supplying the same information such as a password or producing the correct object such as a swipe card which is the same as that earlier stored by the device to represent the genuine user when prompted by the device. The third means of user authentication is by the use of biometrics, in which a person's physiological or behavioural characteristics is used for identification or verification of identity. Keystroke dynamics is a behavioural biometric that utilises a person's habitual way of typing at a terminal such as a computer keyboard or the keypad of a mobile phone to recognise them as demonstrated in [1] and has been proposed as a biometric authentication medium and has also been explored as a means of strengthening user authentication by enhancing the security of password-based authentication on computers and mobile devices [2]. Demographics is one of the variables that may affect the recognition performance of such a biometric modality [3, 4]. Moreover, some biometrics that have been shown to be biased in their performance with respect to different age categories and gender as observed in [5] In literature, our understanding of the interactions between personality traits and the performance of keystroke dynamics authentication on mobile phones has not been sufficiently addressed. This study aims to investigate the impact of age and gender on the performance of a keystroke dynamics-based mobile phone user authentication. The authors in [6] conducted experiments using a desktop computer and keyboard to collect keystroke dynamics data and investigated the uniqueness property of keystroke dynamics. They concluded that there is a significant difference in the recognition performance depending on whether an attacker is a male or a female. In contrast, [7] also investigated the robustness of keystroke dynamics against synthetic forgery attacks using keystroke dynamics data collected on a desktop computer platform but did not observe any significant difference in the performance of their keystroke dynamics classifier between males and female participants. In his thesis, [8], using keystroke dynamics data collected also on a computer platform did not observe any significant influence of the demographic traits considered, including the volunteers age and gender on the tested classifiers' miss rates. A review of relevant literature revealed that while the nature of the relationship between the performance of keystroke dynamics on computer platforms and user demographics such as age and gender has been investigated, not much