Enhanced Network Communication Security Through Hybrid Dragonfly-Bat Feature Selection for Intrusion Detection

Network Intrusion Detection and Prevention Systems (NIDPS) play a critical role in securing network communications by detecting and mitigating cyber threats. Machine Learning (ML)-based NIDPS have proven to be highly effective in identifying network intrusions; however, their performance deteriorates when dealing with high-dimensional data. To address this, an efficient feature selection technique is essential to eliminate redundant or less relevant features, enhancing both accuracy and computational efficiency. A novel hybrid feature selection approach combining the Dragonfly Algorithm (DA) and Bat Algorithm (BA) is proposed to reduce dimensionality. Using the optimized feature subset, classification is performed with Extreme Gradient Boosting (XGBoost), Random Forest (RF), Logistic Regression (LR), and Support Vector Machine (SVM). This study employs the UNSW-NB15 dataset to train and evaluate an NIDPS framework. Experimental results demonstrate that the DAuBA feature selection method significantly improves classification performance, with XGBoost and Decision Tree (DT) achieving 100% accuracy, highlighting the effectiveness of the suggested approach in intrusion detection.