Network intrusion detection using hybrid approach

Abstract

This project presents a new approach to network security by combining two types of detection techniques: signature-based and anomaly-based. The signature-based method helps catch known threats by recognizing attack patterns, while the anomaly detection technique, powered by machine learning (specifically Isolation Forest), identifies unusual or new network behaviors that might signal emerging threats. After rigorous testing with benchmark datasets, the system has shown to be more accurate and generates fewer false alarms than traditional methods. It also includes useful features like storing detected anomalies for later review and sending real-time alerts to ensure prompt responses. This research emphasizes how blending these detection methods can make network intrusion systems more effective, with potential future improvements like integrating real-time monitoring or deep learning for even better performance. The findings are currently being prepared for publication.