Skip to main content

Automated Data Collection for Mobile Network Traffic Data Classification

We are creating an updated dataset mapping mobile network packet metadata to applications, application activities, and application activity types in order to address the problem of mobile network data classification. For example, the dataset we create might contain packet length and proportion of inbound vs outbound packets mapped to instances of streaming videos on Netflix. Such a dataset, in conjunction with robust modeling techniques, could be used by telecommunications companies to dynamically allocate resources to users based on the applications they are using or actions they are performing. The project can be split into three main components: automated data collection, dataset aggregation, and modeling. Our data collection platform collects network packets from autonomous mimicking of user actions on the most popular mobile applications. Relevant data is extracted, transformed, and aggregated from those packets into a usable and robust dataset. We verify the utility of our dataset by training classification models on it and testing on mobile action data collected from real people.

Team Members: 

Neil Charles

Kevin Chau

Christopher Gill

Alex Ma

Jun Min Noh

Jayaj Poudel

Semester