Open Access

ARTICLE

An Intrusion Detection Method Based on a Universal Gravitation Clustering Algorithm

Jian Yu^1,2,*, Gaofeng Yu³, Xiangmei Xiao^1,2, Zhixing Lin^1,2

1 Network Technology Center, Sanming University, Sanming, 365004, China
2 School of Information Engineering, Sanming University, Sanming, 365004, China
3 School of Economics and Management, Sanming University, Sanming, 365004, China

* Corresponding Author: Jian Yu. Email:

Journal of Cyber Security 2024, 6, 41-68. https://doi.org/10.32604/jcs.2024.049658

Received 14 January 2024; Accepted 08 May 2024; Issue published 04 June 2024

Abstract

With the rapid advancement of the Internet, network attack methods are constantly evolving and adapting. To better identify the network attack behavior, a universal gravitation clustering algorithm was proposed by analyzing the dissimilarities and similarities of the clustering algorithms. First, the algorithm designated the cluster set as vacant, with the introduction of a new object. Subsequently, a new cluster based on the given object was constructed. The dissimilarities between it and each existing cluster were calculated using a defined difference measure. The minimum dissimilarity was selected. Through comparing the proposed algorithm with the traditional Back Propagation (BP) neural network and nearest neighbor detection algorithm, the application of the Defense Advanced Research Projects Agency (DARPA) 00 and Knowledge Discovery and Data Mining (KDD) Cup 99 datasets revealed that the performance of the proposed algorithm surpassed that of both algorithms in terms of the detection rate, speed, false positive rate, and false negative rate.

---

Keywords

---