Identification of applications running on smartphones using call detail records

The surge use in network traffic due to increased adoption of 4G LTE-capable portable devices put pressure on network providers to constantly upgrade their network infrastructures. At the same time, unlike traditional Internet access devices (e.g., computer desktops), whose network traffic can be easily tracked via their port numbers, applications of handheld devices typically communicate via HTTP and encrypted HTTPS. Additionally, for scalability purpose, most of their data are sent and received via Content Distribution Networks (CDNs). These communication characteristics obscure any attempt to monitor their network usage. This paper uses cellular network traffic generated by the applications running on a smartphone to predict a 62% prediction accuracy in distinguishing social network services (SNS) from other background applications.