Leveraging Artificial Intelligence Feature Selection for Cybersecurity Network Anomaly Intrusion Detection
Eva-Liisa Kafita, Tatsuya Yamazaki · 2024
With the unremitting prevalent cyberattacks through computer network infrastructures, the development of robust security solutions with improvement requirements of Artificial Intelligence (AI) techniques has become imperative. Such solutions include anomaly intrusion detection systems. By using Feature Selection (FS) ensembled Machine Learning (ML) models and systems, we promptly counter defend cyber threats and malware such as Distributed Denial of Services (DDoS), Brute force, SQL injection, etc. FS enhances ML-based network anomaly intrusion detection models to promptly and accurately classify benign from malware. FS also reduces computing resource requirements and improves detection rate by eliminating excessive noise during data preprocessing, training, and testing stages. Hence, this research aims to contribute to the body of knowledge by presenting and harnessing optimal feature selection process that combines one-class and binary classification ML models’ outputs to propose an ensemble anomaly intrusion detection system. The proposed FS ML model aims to boost prompt cybersecurity measures against network anomaly intrusions. The research used CSE-CIC-IDS2018 dataset to evaluate precision, recall, f1-score and ROC-AUC in determining the performance and success rate. The results indicate that the ensemble model outperformed with an accuracy 0.94, precision 0.95 and ROC-AUC score of 0.96. For this study, the exploration of FS enhanced accurate and precise classification while employing minimum computing resources and time. This method advances Network Intrusion Detection Systems (NIDS) capabilities in overcoming the fast past pace of obsolete systems due to rapid emerging trends of cyber threats and attack vectors, using limited resources yet achieving optimal pick. The Ensemble model enabled us to improve the overall performance and robustness by managing bias-variance trade off.