Towards a Multi-Label Dataset of Internet Traffic for Digital Behavior Classification

Abstract

With the digital transformation of model society, the deep understanding of digital behavior is critical for both users and service providers. Nevertheless this work is challenging due to the lack of an extensive model and the corresponding dataset to support digital behavior classification. In response to this, we presented in this work a complete process of modelling, data collection and classification of user digital behaviors over Internet: firstly the fundamental digital context model is introduced to provide a thorough understanding of digital behavior and digital environment properties. Based on the model, the data collection process is presented and a multi-label dataset of Internet traffic (MLDIT) has been collected with all model properties, finally a first series of classification experiments with MLDIT has been conducted showing promising results to identify user interaction state, applications and actions. Aiming at providing a thorough model of digital behavior and a reference process for data collection and classification, we expect likewise to attract community efforts to collaborate on the MLDIT enrichment.