An unknown Protocol improved k-means clustering algorithm based on Pearson distance

Article type: Research Article

Authors: Yin, Shizhuang^{; 1; *} | Wang, Tao^{; 1}

Affiliations: Army Engineering University Shijiazhuang Campus Hebei Shijiazhuang, China

Correspondence: [*] Corresponding author. Shizhuang Yin, Army Engineering University Shijiazhuang Campus Hebei Shijiazhuang, China. E-mail: 18525742489@163.com.

Note: [1] Shizhuang Yin and Tao Wang contributed equally to this paper.

Abstract: In order to solve the clustering problem of unknown binary protocols, an improved k-means unknown binary protocol clustering method is proposed, which determines the initial clustering center and improves the clustering distance. Firstly, the k value is determined and the clustering center is extracted by using DCBP (Determine the initial clustering center of binary Protocol) algorithm and the change rate of error square, and then the data are clustered by improving the k-means algorithm of distance function. The unknown binary protocol bit stream is divided into different subsets of binary protocols. By improving the k-means algorithm, the Pearson distance improves the accuracy of binary protocol clustering from 96% to 98.9%. The DCBP algorithm helps us to determine the k value accurately. The k value determined in this paper is 5, and the clustering accuracy is 98.9%. The clustering accuracy is 80% when k is 4 and 92.2% when k is 6. And the operation speed of the improved k-means algorithm is better than that of the AGNES algorithm. The algorithm is better adapted to the clustering of unknown binary protocols, and improves the accuracy of clustering and the speed of operation.

Keywords: Protocol identification, unknown binary protocol, Pearson distance, determine cluster center

DOI: 10.3233/JIFS-191561

Journal: Journal of Intelligent & Fuzzy Systems, vol. 38, no. 4, pp. 4901-4913, 2020