Influence Of Fractal Dimension Statistical Charachteristics On Quality Of Network Attacks Binary Classification
Oleg I. Sheluhin, Mikhail Kazhemskiy · 2021
It is proposed to improve network attacks binary classification efficiency by introducing additional statistics of attacks fractal dimension (FD) among other descriptors and compare the performance of several classifiers. The idea of taking into account additional statistical characteristics of network attacks FD is new (in the past, only the average value of the Hurst parameter was considered) and is the main contribution of the article. The effectiveness of the proposed method is shown by evaluating network attacks and normal traffic binary classification quality with machine learning algorithms in case of using the UNSW-NB15 database. Usage of FD distribution average value, variance, skewness and kurtosis coefficients as additional information features that characterize its form and parameters can increase the efficiency of binary classification by an average of 10%.