Інші праці КЕОМ

Permanent URI for this collection

ENG: Other Works

Now showing 1 - 1 of 1

Study of the Configuration of the Neurofuzzy Network to Determine the Degree of Confidence in the Implementation of a DOS attack
(ProConference in conjunction with KindleDP Seattle, Washington, USA, 2024) Pakhomova, Victoria; Govorukha, Yegor
ENG: As a research method, ANFIS configurations 4-5-12-81-81-1 were used, where 4 is the number of input neurons; 5 – total number of layers; 12 – the number of neurons of the first hidden layer; 81 – the number of neurons of the second hidden layer; 81 – the number of neurons of the third hidden layer; 1 – the number of resultant neurons created using the Fuzzy Logic Toolbox of the MatLAB system, the resulting characteristic is the degree of confidence that the DoS attack occurred at the following terms: low; medium; high. Using the open database of NSL-KDD network traffic parameters, a study of the number of terms of input neurons on the created ANFIS1 (three terms each) and ANFIS2 (two terms each) was carried out with the Gaussian function of neuronal membership on samples of different lengths (100, 200 and 300 examples) using different methods of training optimization (Backpropa and Hybrid). It was determined that the smallest values of errors of the first and second kind were three terms for input neurons on the generated ANFIS1 under the Hybrid method.