User authentication via touch pattern recognition based on isolation forest
Alexander I. Filippov, Artem V. Iuzbashev, Alexey S. Kurnev · 2018
The analysis of existing methods of authentication on mobile devices points out that existing approaches either do not provide the necessary level of security, or cause some inconvenience to users. The research suggests one of the ways to improve the security of existing authentication systems while maintaining the convenience of using mobile devices. It introduces continuous user authentication based on the features of their interaction with the device's touch screen. This article investigates the main classes of methods for processing behavioral characteristics applicable to the user authentication task based on behavioral biometrics. The data on the features of interaction with a touch screen was collected from twenty-one users, including the author of this research. The prototype of the authentication system capable of authenticating users quite accurately was obtained as a result of the feature extraction and model fitting based on the isolation forest technique.