DETECTING EMERGING CYBER THREATS WITH AN EFFICIENT HYBRID ENSEMBLE MODEL WITH FEATURE SELECTION

Abstract

By combining a hybrid ensemble learning approach with feature selection, this study offers a real-time cyber threat detection system that is both more efficient and more accurate in its classifications. By using chi-squared feature selection, we can reduce dimensionality and noise in cyber threat intelligence data and extract the most useful textual features. A hybrid ensemble model is trained using the chosen characteristics. This model combines Decision Tree, Extra Tree, and Random Forest classifiers. Reliable and robust prediction for binary and multiclass cyber threat categorization is ensured via a voting mechanism. Deploying the suggested system through a Flask-based web interface allows for interactive and automatic cyber threat analysis on user-provided data, supporting real-time applicability. The suggested extension is useful for detecting both existing and new cyber threats in dynamic contexts, according to experimental results, which show that it greatly improves detection performance and adaptability.