Digital Behavioural Biometrics: A Systematic Review of Reviews

The proliferation of digital data has provided new means to identify people from the way they behave and interact with technologies. This article provides the first systematic review of reviews (n = 41) on digital behavioural biometrics to ascertain what can be inferred about identity from digital sources, and “boundaries” to their applications. On October, 5, 2021, we searched for reviews in the databases PsychInfo, ACM Digital, IEEE Xplore Digital Library, Academic Search Ultimate,Web of Science, and Scopus. We included reviews of digital behavioural biometrics that inferred personal identity, excluding those focused solely on physiological identifiers, personality or health inferences without identity. The 41 reviews document a range of accuracy and error rates across and within behavioural biometric modalities including keystroke, gait, and multimodal which we document in tables and summarise narratively. The research demonstrates that identity can be inferred continuously, cheaply, and covertly, with little user effort, through the development of digital behavioural biometric systems. However, such systems need to be adaptive and consider how behaviour can change over time and context. In particular, the absence of psychological theory in development of the systems leads to suboptimal performance. The reviews also highlight issues and inconsistencies in the way biometric systems are reported alongside a lack of open data. Consequently, we propose a checklist to help researchers follow best practices and to standardise procedures.