AI-Driven Intelligent Firewall for Real-Time Intrusion Detection Using XGBoost and CIC-IDS-2017 Dataset

Network security remains a critical challenge due to increasingly sophisticated cyberattacks. Traditional firewalls employing static rules and signature-based detection prove ineffective against novel and evolving threats. This paper presents an artificial intelligence-driven intelligent firewall system capable of detecting multiple attack types in real-time using machine learning techniques. We preprocessed, cleaned, normalized, and merged network flow data from the CIC-IDS-2017 dataset, specifically utilizing Monday and Wednesday traffic captures. An XGBoost classifier was trained on extracted features, achieving enhanced accuracy in multi-class attack detection. A Flask-based web interface enables real-time CSV traffic prediction with immediate actionable results. The proposed system successfully identifies attacks including Port Scan, Distributed Denial of Service (DDoS), and Denial of Service (DoS) variants, producing labeled outputs for immediate firewall action. Experimental results demonstrate that machine learning-enhanced firewalls significantly outperform traditional rule-based systems in both adaptability and accuracy, validating their essential role in next-generation network security infrastructure.