Performance of EWMA and ANN-based Schemes in Detection of Denial of Service Attack

Abstract

Abstract To ensure successful implementation of cyber-physical systems, industries require computer networks to be protected from malicious attacks. Despite various intrusion detection techniques being proposed by researchers, computer networks are still vulnerable to attacks. As new attacks becoming more complicated, more research is needed to develop more effective and reliable intrusion detection schemes. This study investigated the exponentially weighted moving average control charting technique for detection of malicious denial of service (DoS) trafic and compared it with artificial neural network (ANN) based scheme. Eight features from the Benchmark KDD Cup99 computer network datasets were extracted and their respective ARL1 and false alarm rate were evaluated. The results suggest that EWMA technique is effective only for selective features and the ANN-based scheme is relatively consistent in handling variability in traffic data. This study opens new opportunities for further investigation to enhance performance of the proposed schemes.