A Network Anomaly Detection Algorithm based on Natural Neighborhood Graph

As a kind of network security protection technology, intrusion detection technology has become one of the hot topics in the field of network security. In order to solve the problem that the methods of network anomaly detection have a high requirement on the purity of the normal data-set, and that the existing methods based on outlier detection need to set an anomaly threshold manually. Combining with the idea of Natural Neighborhood Graph, a network anomaly detection method (NAD-NNG) is proposed. In order to eliminate noise points or mislabel points and reduce the time complexity of anomalies detection, the algorithm uses the Natural Neighborhood Graph to cluster the normal data-set. Also, the algorithm can adaptively obtain a percentage value β for setting the anomaly threshold. Experiments on KDDCUP99 show that compared with the other two algorithms, the proposed method can achieve a higher detection rate based on a tolerable false alarm rate.