Packet Classification using Support Tensor Machines

Antonio Augusto Teixeira Peixoto, Carlos Alexandre Fernandes, Laise S Santos

DOI: 10.14209/sbrt.2022.1570825020

Evento: XL Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2022)

Keywords: packet classification tensor support tensor machine

Abstract

A wide variety of packet classification algorithms exist in the research literature and commercial market. The existing solutions exploit various design tradeoffs, providing high search rates, power and space efficiency and the ability to scale to large numbers of filters. However, still remains a need for techniques that achieve a favorable balance among these tradeoffs and scale to support classification. Based on this motivations, this paper presents a tensor approach for the classification of TCP and UDP packets. By using a multidimensional structure, more specifically a 4-th order tensor, to store the packet data, a tensorial algorithm known as Support Tensor Machines (STM) is used to perform classification. Results showed good performance of the approach in comparison to other classifiers such as the Support Vector Machines and Naive-Bayes.

Download