GRU-Based Learning for the Identification of Congestion Protocols in TCP Traffic

Abstract

This paper presents the identification of congestion control protocols TCP Reno, TCP Cubic, TCP Vegas, and BBR on the Marist University campus, with an accuracy of 97.04% using a GRU-based learning model. We used a faster neural network architecture on a more complex and competitive network in comparison to existing work and achieved comparably high accuracy.