TCM-KNN scheme for network anomaly detection using feature-based optimizations

Yang Li, Li Guo

Abstract: With the rapid increase of network threats and cyber attacks, network security problem is becoming more and more serious. Network anomaly detection is a key technique to secure information systems and resist cyber attacks. In this paper, we first propose an efficient network anomaly detection technique based on TCM-KNN scheme. Secondly, we emphasize the feature-based optimizations for our TCM-KNN. We employ feature selection and feature weight mechanisms to optimize TCM-KNN as a promising lightweight and on-line anomaly detection technique both in reducing its computational cost and in boosting its detection performance. A series of experiments on well-known intrusion detection dataset KDD Cup 1999 demonstrate the effectiveness of our methods presented in this paper.