Abstract
Network security threats continue to escalate with increasingly sophisticated intrusion attempts targeting both enterprises and critical infrastructure. Traditional signature-based intrusion detection systems (IDS) struggle with high false positive rates and inability to detect zero-day and mutation-based attacks. This paper presents NIDS Sentinel, an AI-driven Network Intrusion Detection System leveraging ensemble machine learning models for real-time cyber threat detection. The system integrates three supervised learning algorithms—Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Multi-Layer Perceptron (MLP) Neural Networks—trained on NSL-KDD-style network traffic datasets. Our comprehensive approach includes data preprocessing with feature normalization, multi-model evaluation, and a production-grade three-tier architecture combining Python ML backend, Node.js/Express API middleware, and React.js web dashboard. Experimental evaluation demonstrates 96.9% overall detection accuracy across all models, with XGBoost achieving 97.1% individual accuracy while maintaining low false positive rates. The system successfully identifies diverse attack types including Denial of Service (DoS), Remote-to-Local (R2L), User-to-Root (U2R), and Probe attacks with real-time processing capability. A live monitoring dashboard enables security analysts to upload datasets, visualize predictions, and track threat patterns without requiring deep machine learning expertise. This work validates the effectiveness of ensemble ML approaches combined with modern web technologies for practical, deployable intrusion detection systems.