EXPERIMENTAL STUDY OF THE KERNEL DENSITY IN THE ANGELOV-YAGER FUZZY SYSTEMS FOR SOLVING CYBERSECURITY PROBLEMS
M.O. Svetlakov, K.S. Sarin, M.B. Bardamova, N.P. Koryshev, R.E. Kolomnikov, Y.A. Shurygin, I.A. Hodashinsky
Tomsk State University of Control Systems and Radioelectronics, Tomsk, Russia
Keywords: Angelov-Yager fuzzy systems, kernel density functions, cybersecurity, streaming data, classification, clustering
Abstract
Effective elimination of cyber threats involves detecting attacks and taking protective measures in real time based on streaming data, but existing solutions are not always able to process data in this mode. To solve these problems, it is proposed to use the incremental learning method - fuzzy classification and clustering systems of the Angelov-Yager first-order type. The original implementation of the system uses the Cauchy kernel function; the use of alternative kernel functions has not been considered before. An experimental study of alternative kernel functions on cybersecurity data sets is conducted, a statistically significant difference in the accuracy and number of rules in the application of various functions is revealed, and the best kernel functions for individual data sets are identified
|
