Naïve Bayes Classifier and Particle Swarm Optimization Feature Selection Method for Classifying Intrusion Detection System Dataset

A. S. Talita, O. S. Nataza, Z. Rustam

Research output: Contribution to journalConference articlepeer-review

21 Citations (Scopus)

Abstract

The security of a network might be threatened by an intrusion aim to steal classified data or to find weaknesses on the network. In general, network main security systems use a firewall to control and monitor both incoming and outgoing network traffic. Intrusion Detection System can be used to strengthen network security. Several data mining methods have been used to solve Intrusion Detection System (IDS) problem on a network. On this paper we will use Naïve Bayes Classifier along with Particle Swarm Optimization (PSO) as the feature selection method specifically on one of the benchmark dataset on IDS problem, KDD CUP'99. The dataset consists of more than 40 features with more than 400 thousands records. To solve IDS problem on the dataset, it needs a quite expensive cost either on time computation or memory usage hence the use of PSO as the feature selection method. The best classification result was reached when we use 38 features where the accuracy is 99.12%. Particle Swarm Optimization method has several parameters that may affect the classification performance. For future improvement, it is possible to use a parameter optimization method to ensure the best classifier performance.

Original languageEnglish
Article number012021
JournalJournal of Physics: Conference Series
Volume1752
Issue number1
DOIs
Publication statusPublished - 15 Feb 2021
Event3rd International Conference on Statistics, Mathematics, Teaching, and Research 2019, ICSMTR 2019 - Makassar, Indonesia
Duration: 9 Oct 201910 Oct 2019

Keywords

  • Classification
  • Intrusion Detection
  • Naïve Bayes
  • Particle Swarm Optimization

Fingerprint

Dive into the research topics of 'Naïve Bayes Classifier and Particle Swarm Optimization Feature Selection Method for Classifying Intrusion Detection System Dataset'. Together they form a unique fingerprint.

Cite this