Enhancing Network Intrusion Detection Systems Through Dimensionality Reduction

The proliferation of data on the Internet has intensified the difficulties faced by network intrusion detection systems (NIDS) in handling large dimensions of data that include irrelevant and duplicate elements. This process entails a significant investment of time and effort in accurately identifying the attack while also experiencing a rise in false alarms. Applying dimensionality reduction may solve this problem. This paper employs and compares two feature selection methods, the Whale Optimization Algorithm (WOA) and Hawks Optimization (HHO). These methods reduce the number of features by finding the minimum number without affecting the performance of the NIDS system. The main idea is to select the key features from the NSL-KDD datasets using WOA and HHO. The results have shown a better performance of the NIDS system when using HHO rather than WOA with binary classification. The HHO and WOA achieved 93.83% and 92.83% accuracy, respectively. These results were achieved using the Support Vector Machine Classifier (SVM).