FEBA: An Action-Based Feature Extraction Framework for Behavioural Identification and Authentication

While the usage of behavioural features for authentication purposes is gaining more and more consensus in the community, there is less consensus on which specific behavioural traits may be useful in eventually different settings. This calls for flexible tools which the application developer can leverage to automate the extraction and management of behavioural features for identification and authentication. This paper specifically describes a framework called FEBA (Feature Extraction Based on Action), which to the best of our knowledge is the first open-source framework that provides the developer with simple and flexible means to: i) define application-specific actions, ii) recognize actions based on the received raw data, and iii) finally extract the action-specific features. We have built a complete implementation of FEBA, and made it available online to facilitate future research in such context. To prove the performance of FEBA, we provide an experimental evaluation of a use case scenario, i.e., mouse movements feature extraction and pattern recognition. We believe that FEBA will help researchers and developers to design and implement novel behavioural authentication mechanisms.