Authors - Nithya B, V.Venkataraman, Nithin Balaaji D V, Chandra Chud Abstract - Transmission Control Protocol is widely popular with varied applications in internet like World Wide Web, email, remote administration and file transfer protocols. Over the years, numerous congestion control algorithms for TCP have been proposed each with its own characteristics. The automatic identification of individual algorithm is important for traffic engineering in the Internet. This paper aims to classify three major congestion control algorithms : TCP New Reno, TCP High Speed and TCP Veno. The proposed approach passively collects packet trace files with data and ACK segments from an NS3 simulator and then calculates a time sequence split of the congestion window. The collected data are normalized and fed separately to machine learning approaches: CNN, LSTM, hybrid CNN-LSTM and a ConvLSTM. The performance of each of these approaches is analyzed in terms of accuracy, precision, recall and F1 score. From the results, it is inferred that the hybrid CNN-LSTM model gives highest accuracy in classifying TCP congestion control algorithms.
Open Zoom