An anomaly detection model based on neighborhood preserving

摘要: Real-time detection is one of the most important issues in intrusion detection. When network data becoming huge and high dimensionality, real-time with accuracy low false alarm rate challenging for previous methods. In order to handle this problem appropriately, we propose a novel anomaly model based on neighborhood preserving branch bound tree, named as ADM-NP. method, high-dimensional mapped lower dimension space, reducing complexity computation algorithm. Besides, since tree adopted, KNN searching algorithm becomes faster. Empirical studies KDD CUP 99 set demonstrate effectiveness our method.