Advanced Cyber Security Threat Intelligence with Hybrid Predictive Model Integration

Abstract

This extended study presents an enhanced cyber-attack detection framework using a hybrid stacked model that integrates Multilayer Perceptron (MLP), K-Nearest Neighbor (KNN), and Random Forest (RF) classifiers. The stacked ensemble combines the strengths of these base models to deliver high robustness and accuracy across heterogeneous benchmark datasets including NSLKDD, CICIDS2017, CICDDOS2019, and X-IIOTID. The system is deployed through a Flask-based web interface, enabling real-time cyber-attack prediction and user interaction with uploaded test datasets. Incorporating Explainable AI (XAI) techniques provides interpretability by highlighting key feature contributions, supporting transparent and informed decision-making. Experimental results demonstrate superior performance, achieving up to 100% accuracy, proving the efficiency of the hybrid stacked framework as a scalable and intelligent solution for proactive cybersecurity defense.