Abstract—The swift expansion of intricate and multifaceted cybersecurity data presents significant obstacles for effective and precise threat identification. An Enhanced Support Vector Machine–Cluster-based Recursive Feature Elimination (E-SVM–CRFE) framework that emphasizes adaptive and data-driven feature selection is presented in the paper as a solution to this problem. In order to identify statistical, behavioral, and pattern-based indications of cyber dangers, the suggested approach starts with thorough data pretreatment and multi-perspective feature extraction. A similarity-based clustering technique is used to group features that show high correlation or redundancy, and representative features are chosen based on their SVM-derived significance weights. The feature subset is iteratively refined by an adaptive elimination method that balances dimensionality reduction and classification performance under the guidance of dynamic thresholding. When compared to conventional SVM-based intrusion detection systems, experimental evaluations show that E-SVM–CRFE greatly improves detection accuracy, lowers computing cost, and promotes the identification of uncommon and complicated attack behaviors. The outcomes demonstrate how adaptive feature refinement can enhance model interpretability and generalization. Keywords— cybersecurity, Support Vector Machine, Cluster, Recursive Feature Elimination and data pre-processing.
Brindha et al. (Sun,) studied this question.