Convolutional feature extraction from human computer interaction data in behavioural biometric
Abstract
The paper considers methodology of user biometric profile formation from human computer interaction (HCI) data by applying a discrete-time convolution to set of measured user activity parameters. Presented technology was successfully verified in pilot study aimed at determining effective approaches to the user behavior analysis problems. Provided detailed experiment design description and formal specification of initial parameters used for building up behavioral profiles. Obtained results allow to conclude that certain parameters emerging in typical HCI scenarios could be used to provide a mechanism for the “continuous” user authentication procedure after standard authorization was performed.
