Vol.26, No.4, 2020, pp.795-805, doi:10.32604/iasc.2020.010114
A Perspective of the Machine Learning Approach for the Packet Classification in the Software Defined Network
  • B. Indira1,*, K. Valarmathi2
1 Department of Computer Science and Engineering, P.S.R. Engineering College, Sivakasi-626 140
2 Department of Electronics and Communication Engineering, P.S.R. Engineering College, Sivakasi-626 140
* Corresponding Author: B. Indira, aridnib@gmail.com
Packet classification is a major bottleneck in Software Defined Network (SDN). Each packet has to be classified based on the action specified in each rule in the given flow table. To perform classification, the system requires much of the CPU clock time. Therefore, developing an efficient packet classification algorithm is critical for high speed inter networking. Existing works make use of exact matching, range matching and longest prefix matching for classification and these techniques sometime enlarges rule databases, thus resulting in huge memory consumption and inefficient searching performance. In order to select an efficient packet classification algorithm with less memory consumption and high classification accuracy, Machine Learning (ML) algorithms are used. For performance comparison, ML algorithms are used, namely Multi-layer Perceptron (MLP), K-Nearest Neighbor (KNN), Decision Tree (DT), Random Forest (RF), AdaBoost classifier (AB) and Support Vector Machine (SVM). All these algorithms build network for packet classification and train the network with the use of Access Control List (ACL) netbench dataset. 5-features of IPv4 packet header are used and the algorithms classify the packets based on action/flow of each packet. Experimental results show that among six algorithms, RF algorithm gives better improvement in accuracyperformance for permitted packets.
Machine learning algorithm, min-max normalization, IPv4 packet header, packet classification.
Cite This Article
B. Indira and K. Valarmathi, "A perspective of the machine learning approach for the packet classification in the software defined network," Intelligent Automation & Soft Computing, vol. 26, no.4, pp. 795–805, 2020.
This work is licensed under a Creative Commons Attribution 4.0 International License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.