A Network Anomaly Detection Method Based on Relative Entropy Theory

Abstract

Network anomaly detection technology has been the research hotspot in intrusion detection (ID) field for many years. However, some issues like high false alarm rate, low detection rate and limited types of attacks which can be detected are still in existence so its wide applications in practice has been restricted. A new network anomaly detection method has been proposed in this paper. The main idea of the method is network traffic is analyzed and estimated by using Relative Entropy Theory (RET), and a network anomaly detection model based on RET is designed as well. The numerical value of relative entropy is used to alleviate the inherent contradictions between improving detection rate and reducing false alarm rate, which is more precise and can effectively reduce the error of estimation. On the 1999 DARPA/Lincoln Laboratory IDS evaluation data set, the detection results showed that the method can reach a higher detection rate at the premise of low false alarm rate.