Behavioral Biometrics in the Era of Artificial Intelligence

Abstract

Biometric technologies have historically been explored as Pattern

Recognition systems. Over the past three decades, biometric- based human recognition systems have significantly changed

and improved. Applications in forensics, surveillance, healthcare, automotive, and human-computer interaction have benefited of such advancements and are currently globally used. However,

the exclusive focus on pattern recognition may obscure or re- strict the potential and capabilities of this discipline. Emerging

biometric modalities have begun to impact the security of sensi- tive data, information, and systems. As the biometric challenges

increase, the solution strategies shifted the attention on human behavior. Behavioral biometrics is the study of patterns in human activities that can be uniquely identified and measured. Recent technological advancements, especially in Artificial Intelligence, as well as hardware development, have increased the potential of biometric approaches and expanded their application fields. Beginning with the wide concept of behavioral biometrics, this thesis aims to advance the state-of-the-art in several applications, such as estimating an individual’s head rotation to determine its intent or attention and analyzing facial expressions to detect

human emotional state. Finally, behavioral biometric traits are in- vestigated through users’ touch interactions with modern mobile

devices. For each of the presented methods, the complete pro- cessing pipeline is described, including data acquisition, feature

extraction, experimental protocols, and decision-making, as well as a comparison to state-of-the-art methods to show advantages and discuss current challenges. [edited by Author]