The XGBoost Model for Network Intrusion Detection Boosted by Enhanced Sine Cosine Algorithm

Network intrusion detection systems are created with the purpose of detecting and identifying threats and vulnerabilities of a target network. One of the most cardinal challenge with such systems is that they frequently generate a considerable high rate of false positives and false negatives and that has a profound influence on its efficiency. However, this issue can be mitigated by applying machine learning models and algorithms. Therefore, in the research shown is in this paper, drawbacks of network intrusion detection systems are addressed by applying well-known XGBoost classifier tuned with improved sine cosine metaheuristics. It is known that machine learning models such is XGBoost should be optimized for each practical problem (dataset) and in this study was shown that XGBoost hyper-parameters' tuning can be efficiently conducted by applying improved sine cosine algorithm. Proposed hybrid framework is validated against benchmarking NSL-KDD dataset and compared to XGBoost without tuning, as well as with few other metaheuristics approaches used for XGBoost tuning, including original sine cosine algorithm. Obtained research findings prove that proposed method is able to improve classificthe position of the current solution that is shown ation accuracy and precision compared to other methods including in the analysis.