An exploration of dynamic biometric performance using device interaction and wearable technologies

With the growth of mobile technologies and internet transactions, privacy issues and identity check became a hot topic in the past decades. Mobile biometrics provided a new level of security in addition to passwords and PIN, with a multitude of modalities to authenticate subjects. This thesis explores the verification performance of behavioural biometric modalities, as previous studies in literature proved them to be effective in identifying individual behaviours and guarantee robust continuous authentication. In addition, it addresses open issues such as single sample authentication, quality measurements for behavioural data, and fast electrocardiogram capture and biometric verification. The scope of this project is to assess the performance and stability of authentication models for mobile and wearable devices, with ceremony based tasks and a framework that includes behavioural and electrocardiogram biometrics.

The results from the experiments suggest that a fast verification, appliable on real life scenarios (e.g. login or transaction request), with a single sample request and the considered modalities (Swipe gestures, PIN dynamics and electrocardiogram recording) can be performed with a stable performance. In addition, the novel fusion method implemented greatly reduced the authentication error.

As additional contribution, this thesis introduces to a novel pre-processing algorithm for faulty Swipe data removal. Lastly, a theoretical framework comprised of three different modalities is proposed, based on the results of the various experiments conducted in this study. It's reasonable to state that the findings presented in this thesis will contribute to the enhancement of identity verification on mobile and wearable technologies.