The increasing popularity of smartphones has raised serious safety concerns. This is due to the fact that these devices hold sensitive personal and often pro fessional information and existing authentication schemes have proven inefficient. Password patterns and PIN codes, in particular, can easily be acquired by attackers with shoulder surfing techniques, while all widely-employed user authentication mechanisms, in general, offer one-time authentication, leaving the device unpro tected after the login stage. In this thesis, a continuous and implicit authenti cation (CIA) approach is introduced that can act as a complementary authenti cation method. This approach is supplemented by developing a methodology of personalising authentication criteria by analysing how different users behave based on the context of the screen they are browsing. This last addition serves as the greatest contribution of this thesis in the field of continuous and implicit authentication, since not many ways of optimizing authentication schemes have been explored yet. As a means of pursuing the aforementioned goals, a behavioral biometrics dataset, containing several users’ gestures, was utilized. Two types of gestures were examined, swipes and taps, on how they can serve as a way of distinguishing users. One-Class SVM played a key role in developing this methodology as it allows training with the use of only one user’s gestures, something that can be deployed in real-life scenarios. The problem of determining the behavioral variance that each user indicates (based on the context of the screen he/she is browsing) was handled as a clustering problem, addressed by the k-means algorithm. The method proved to be efficient, especially when analysing swipe gestures, and the incorporation of contextual-behavioral information can offer substantial improvements in user authentication schemes.