ENHANCING NETWORK INTRUSION DETECTION SYSTEMS USING ENSEMBLE LEARNING AND FEATURE SELECTION TECHNIQUES

ABSTRACT

With the escalating frequency and sophistication of cyber threats, the need for robust network intrusion detection systems (NIDS) has become paramount. This research focuses on enhancing the performance of NIDS through the application of ensemble learning techniques and advanced feature selection methods. The study leverages a comprehensive dataset obtained from [dataset source] to conduct a systematic evaluation of various classification algorithms, including Random Forest, AdaBoost, Gradient Boosting, and others. Additionally, a Filter-Wrapper approach employing ReliefF and Random Forest is implemented for feature selection. The combined methodology is demonstrated to significantly improve the accuracy and efficiency of NIDS. Key findings reveal that ensemble learning, particularly Random Forest, emerges as a standout classifier, exhibiting superior performance in detecting network intrusions. The integration of feature selection techniques substantially reduces computational overhead while preserving high accuracy levels. This research not only contributes to the advancement of network security but also holds practical implications for real-world deployment. The optimized models offer resource-efficient solutions, potentially reducing hardware requirements for NIDS. Future research avenues encompass exploring adaptability in dynamic adversarial environments and integrating deep learning architectures for enhanced anomaly detection. Overall, this study provides a substantial leap forward in the field of network intrusion detection, presenting a solid foundation for further research endeavours aimed at fortifying digital ecosystems against evolving cyber threats.