Mobile apps classification with risk score by exploiting the enriched information of app context

Abstract

With the use of mobile devices increasing rapidly day by day, huge numbers of mobile apps are coming into the market, many of which provide same functionality; as a result, having a proper classification of these apps can be useful for various purposes like making it time efficient and easy to the user for selecting the required app, understanding the user preferences which can motivate the intelligent personalized services, etc. But for having proper mobile app usage analysis, effective classification of apps is required for which detailed information about the apps is needed. However this is a nontrivial task as limited contextual information is available. As the information available about the apps is short and sparse, the classification of these apps can also be considered as coming in the category of classification of short and spares text. To classify these short and spares text, various methods are present that can be used to classify the mobile apps. In this paper, we have presented a method in which we extract the information about the apps from the sources like information from the labels (app name), information from the web search engine (snippets), contextual usage logs of users and the permissions the app requests before installation. This gives an effective and secure classification of the apps as source for most of these apps are some unknown vendors and so they are having the higher possibility of being malicious. With the contextual information collected the designed system is able to recommend the apps to the user based on their preferences.