A Novel Unsupervised Anomaly Detection Approach for Intrusion Detection System

Network Anomaly Detection plays an important part in network security. Among the state-of-the-art approaches, unsupervised anomaly detection is effective when dealing with unlabelled data. However, these approaches also suffer from high false positive rate. We observed that different methods have their own defects and advantages. Inspired by this observation, we provide a new ensemble clustering(NEC) method to detect novel anomalies. In our system, we can get higher detection rate and lower false positive rate compared with existed apporaches as verified over NSL-KDD 2009 dataset.