Cyber Attacks Detection Using Machine Learning Algorithms

This research focuses on the effect of the genetic algorithm in the improvement of machine learning models for NID by using the CICIDS2022 data set. The routing research problem that has been primarily focused is related to the increase in classification accuracy and the optimization of the cyber security systems using intelligent methods of feature selection along with the tuning of the classification models. We ran Random Forest (RF) and Support Vector Machine (SVM) to assess a better predictive accuracy, precision, recall, and running time on each case. The data set with a total of 15031 instances was used and divided into training and test set with a ratio of 80:20 and the results have been analyzed with standard metrics along with confusion matrix analysis. The results depict that with the application of GA in RF and SVM both the outcomes were `RF with GA scored a higher accuracy of 99.30% when compared to standard RF with 99.27% and without GA in SVM 98.97% while with GA, it increased to 99.00%. Analysis of the confusion matrix showed less disparity in the GA variants of the methods. However, the time taken for the processing was high especially for SVM + GA. The results can be generalized as observing that with GA, accuracy is slightly higher than then obtained with P0 but the computational cost is considerably high. It is deduced that GA with RF is the most efficient optimization model in terms of both performance and efficiency.