Intrusion Detection and Performance Analysis Using Copula Functions

Abstract

Nowadays, interest in technology is growing as technology advances and makes our jobs easier. These rapid technological advancements bring with them a slew of unwanted negative attacks, such as cyber-attacks and unauthorized access. To prevent such negative attacks, intrusion detection systems are frequently used. In this research, we make some suggestions for novel and reliable classifiers for intrusion detection systems that are based on copulas. Using copula-based classifiers, we hope to detect intrusion in computer networks. Student's-t, Gumbel, Clayton, Gaussian, Independent and Frank classifiers, which are frequently used in the literature, have been preferred as copula-based classifiers. These classifiers were used to perform classification on the KDD'99 dataset. The 10-fold cross-validation method has been used in the classification phase. When the experimental results were examined, the proposed Gaussian copula-based classifier outperformed state-of-the-art basic methods on the KDD'99 dataset with a success rate of 99.41%. As a direct consequence of this, classifiers based on the copula have shown promising results in the field of intrusion detection. Classifiers that are based on the copula have been found to be a competitive alternative to the most recent and cutting-edge fundamental approaches.