Open AccessSelection of Online Network Traffic Discriminators for on-the-Fly Traffic Classification
Author(s) -
Angela M. Vargas-Arcila,
Juan Carlos Corrales,
Álvaro Rendón Gallón,
Araceli Sanchís
Publication year - 2021
Publication title -
revista ingenierías universidad de medellín/revista ingenierías universidad de medellín
Language(s) - English
Resource type - Journals
eISSN - 2248-4094
pISSN - 1692-3324
DOI - 10.22395/rium.v20n38a4
Subject(s) - computer science , traffic classification , selection (genetic algorithm) , data mining , relevance (law) , set (abstract data type) , context (archaeology) , process (computing) , traffic analysis , traffic flow (computer networking) , domain (mathematical analysis) , machine learning , artificial intelligence , quality of service , computer network , paleontology , mathematical analysis , mathematics , political science , law , biology , programming language , operating system
There are several techniques to select a set of traffic features for traffic classification. However, most studies ignore the domain knowledge where traffic analysis or classification is performed and do not consider the always moving information carried in the networks. This paper describes a selection process of online network-traffic discriminators. We obtained 24 traffic features that can be processed on the fly and propose them as a base attribute set for future domain-aware online analysis, processing, or classification. For the selection of a set of traffic discriminators, and to avoid the inconveniences mentioned, we carried out three steps. The first step is a context knowledge-based manual selection of traffic features that meet the condition of being obtained on the fly from the flow. The second step is focused on the quality analysis of previously selected attributes to ensure the relevance of each one when performing a traffic classification. In the third step, the implementation of several incremental learning algorithms verified the usefulness of such attributes in online traffic classification processes.