Supervised machine learning methods for cyber threat detection using genetic algorithm

Security threats continue to pose enormous challenges to network and applications security, particularly with the emerging IoT technologies and cloud computing services. Current intrusion and threat detection schemes still experience low detection rates and high rates of false alarms. In this study, an optimal set of features were extracted from CSE-CIC-IDS2018 using genetic algorithm. Machine learning algorithms, including random forest, support vector machines, logistic regression, gradient boosting, and naïve bayes were employed for classification and the results compared. Evaluation of the performance of the proposed cyber security threat detection models found random forest as the highest attacks detection with 99.99% accuracy. K-nearest neighbor achieved 99.96% while a detection accuracy of 97.39% was obtained by support vector machines. The model which used gradient boosting obtained an accuracy of 99.97%, and the logistic regression model achieved a 94.94% accuracy. The lowest accuracy rate was obtained by the naïve bayes model with a detection accuracy of 68.84%.