Behaviour Biometrics Using AI for Continuous Authentication Systems

Behaviour-based continuous authentication systems like those that utilise an individual user's typing rhythm and device usage behaviour patterns have much potential over password-based schemes since they do not require an individual user to memorise passwords. Irrespective of the progress in biometric technologies, many systems are still susceptible to more complex attacks, and the decision between security and usability has been a perennial struggle for researchers and practitioners. This paper offers a powerful continuous authentication system based on AI and behavioural biometrics, enhancing precision and resistance to motivated attacks. By measuring behavioural data (e.g. keystroke dynamics and motion sensor) on a heterogeneous user population and then analyzing it, we trained our machine learning models to verify users in real-time. Based on our results, our performance is better than that of the traditional and static authentication methods, with an accuracy of 97.2 per cent, with the false acceptance rate (FAR) and false rejection rate (FRR) of 1.8 per cent and 2.3 per cent, respectively. Moreover, error analysis depicted significant trends in behaviour changes, which apply to an adaptive security strategy. This is demonstrated as a potential of AI-based behavioural biometrics to support feasible, secure, and user-friendly continuous authentication systems that work in contemporary cybersecurity scenarios. Future improvements will additionally involve enlarging datasets, combining multi-modal behavioural features, and increasing resistance to spoofing and behavioural drift.