USING DECISION TREE AND MACHINE LEARNING TO RECOGNIZE USERS BY THEIR BEHAVIOUR

This research proposes a complementary method for authenticating users on mobile applications. The method uses decision trees to recognize user behaviour based on specific measurable properties such as the speed of text typing, the preferred zoom factor, the reading and writing mode or the preferred mode of closing the keyboard. By continuously doing measurements on these properties every time the user is using the application, a user profile can be generated, stored and used by the application to authenticate the rightful owner. This functionality can be further extended using Machine Learning algorithms such as Sofimax Regression to recognize a specific user from a given group.