An Approach of Role Updating in Context-Aware Role Mining

Introduction

Recent years have witnessed a rapid development in Internet of Things (IoT) and mobile technologies. IoT refers to the internetworking of physical devices, vehicles, buildings and other items embedded with electronics, software, sensors, actuators, and network connectivity that enable these objects to collect and exchange data (ITU, 2016). With the objective of allowing people and things to be connected anytime, anyplace, with anything and anyone, ideally using any path/network and any service (Sundmaeker, 2010), IoT will have a tremendous impact on the way people work and live. With the prevalence of IoT, an increasing number of real-world things will be connected to the IoT environment by exposing their features and capabilities as services, which will lead to a proliferation in services (Huang, 2016; Zhang, 2012). Similarly, as an essential enabling technology of IoT, the popularity of mobile technologies also stimulates the bloom of available services in the app stores such as Apple AppStore and Google Play.

Context-aware communication and computing is a key technology that enables intelligent interactions what the IoT paradigm envisions (Perera, 2014). Services provided by the pervasive and mobile environments of IoT are context-aware in essence. The increasing capability of collecting context information of users, e.g., location, time, and even users’ current status, offers plenty of opportunities for service providers to provide personalized services for their customers, e.g., to make these devices more intelligent and adaptive to user behaviors by automatically recommending suitable services (Adomavicius, 2011; Woerndl, 2009; Tian, 2015). This is especially important due to the increasing number of services available in the IoT and mobile environments.

Service recommendation has been a hot topic in recent years. Most existing service recommendation approaches are based on individual intelligence or the local knowledge (e.g., similar neighbors) of users (Su, 2009), while neglect the common knowledge among the user community. In (Wang, 2012), the notion of “roles” is adopted to represent user groups, which represent the abstract characterization of user behaviors within a certain context. Users can play different roles in their daily life, and their roles change dynamically as the context changes. Users who play the same role are likely to share the same preferences and behavior patterns. Once a user’s role in a given context is identified, the services closely related to the role can be recommended to the user. The services recommended by those who play the same roles as the user are usually more reliable or acceptable than those recommended by those who seldom play the same roles. However, users’ roles are usually not explicitly specified; therefore, how to extract users’ roles from users’ behavior patterns becomes a key issue.

Many works have been conducted to find effective ways of context aware role mining (Wang, 2012; Wong, 2012; Wong, 2014; Hong, 2015). Given a three-dimension matrix <user, context, behavior>, these approaches can be used to generate a user-role assignment matrix as well as a role-behavior assignment matrix. In this way, the latent roles in the dataset can be identified. In these approaches, the input data set is usually fixed. However, in the pervasive IoT environment, the new coming user behavior patterns and the behaviors of existing users under various contexts might change frequently, which will result in the update of mined roles. Most existing works on role mining neglect the dynamic change of the input data set and how to update the extracted roles when user behavior patterns change have not been sufficiently discussed. Towards this issue, in this paper, the authors propose a novel approach to update roles by adapting the existing results of role mining rather than re-mining the whole dataset when the user behavior patterns change. More specifically, several algorithms are proposed towards various scenarios such as new users and new contexts are removed from and added into the input data. Experiments on a simulated data set show that compared with the algorithm that takes all the updated data as the input, our approach can significantly decrease the update time. In addition, compared with a role updating algorithm in (Wong, 2014), ours can guarantee the completeness of the extracted roles in the updating process.

The structure of the paper is organized as follows. First, the authors discuss related works. Then the definition of context-aware role mining and the role updating problem are presented. Afterwards, the proposed role updating algorithms are given, followed by detailed experimental analysis. Finally, the authors conclude the paper.