DESIGN AND DEVELOPMENT OF AN AI-DRIVEN NETWORK DEFENSE SYSTEM FOR REAL-TIME THREAT DETECTION, PREVENTION, AND TRAFFIC ANALYSIS

This project presents the design and development of an AI-driven network defense system for real-time detection, prevention, and analysis of cyber threats. Traditional security systems struggle with advanced and evolving attack types, creating the need for adaptive solutions. A virtual enterprise network was built in EVE-NG, incorporating VLAN segmentation and simulated attack scenarios from a Kali Linux node. Real-world traffic from the CICIDS2018 dataset was used to train a Random Forest Classifier, with preprocessing steps including feature selection, scaling, and dimensionality reduction. The trained model was deployed within a Python-based detection script using PyShark, enabling live traffic capture, classification, alerting, and automated blocking of malicious IPs. Testing showed the model achieved 89% accuracy with strong F1-scores across most attack categories maintaining an average low false positive rate of 2%. The system successfully detected and blocked both internal and external threats within the simulated environment with minimal response delay. Unlike prior works focused on offline detection, this project demonstrates a practical proof-of-concept for AI-driven intrusion prevention, showing that machine learning can enhance enterprise network defense by providing smarter monitoring, faster response, and more adaptive protection.