Evaluation system for user authentication methods on mobile devices

Abstract

Nowadays, special interest is taken on transparent and reliable user authentication on mobile devices. Variety of proposed authentication methods allows developers to choose flexibly authentication methods according to predefined accuracy, usability and software (hardware) requirements. Nevertheless, implementation of these methods in real devices requires tedious performance analysis of proposed solutions in real situations. The analysis is complicated due to lack of open implementation of proposed authentication methods and results of preliminary testing on different datasets under various metrics.In this work we propose the universal system for evaluation of user authentication methods. The case of authentication on mobile devices, such as smartphones, is considered. The proposed system admits smooth processing of various user features databases without usage of specific importer modules. Imported features are analyzed for reconstruction of multimodal (behavioral) user models that can be used in advanced behavior-based authentication methods. The wide range of feature processing methods and performance indices, such as FAR, FRR, etc., is supported that allows developers to improve authentication method on the fly without re-building of the prototype.