A Multi-Expert Based Approach to Continuous Authentication of Mobile-Device Users

Abstract

Currently used in mobile devices PIN-based user authentication cannot provide a sufficient security level. Methods based on multi-modal user authentication involving biometrics (i.e. physical and behavioral characteristics of a person) may be employed to cope with this problem. However, dealing with physical characteristics only, these methods are either unable to provide continuous and user-friendly identity verification, or are resource consuming.In this paper, we aim at the provision of continuous, user-friendly and accurate verification of the user identity while preserving scarce resources of a mobile device. Rather than physical, behavioral characteristics are analyzed. The normal behavior of the user is modeled by a set of complementary behavioral aspects that can be used to uniquely identify the user. We develop an approach, where these aspects are separately monitored by dedicated software-based experts. By analyzing the deviations of the current behavior from the modeled one, each expert infers separately its decision about the user identity. The final decision is derived from these multiple expert decisions by applying a decision fusion technique. The monitoring of multiple behavioral aspects helps to improve the authentication accuracy and enables the continuous verification of the user identity. The user-friendliness is supported by the use of transparent authentication methods that do not require direct user participation. Finally, the analysis of behavioral characteristics and the decision fusion process do not involve complicated computational steps and, therefore, are conservative in resource consumption.Keywordsmobile device securityauthenticationcontinuous identity verificationuser profilingexpert decision fusion