Network Traffic Analysis Using Refined Bayesian Reasoning to Detect Flooding and Port Scan Attacks

Abstract

Dynamical analysis of the current network status is critical to detect large scale intrusions and to ensure the networks to continually function. Collecting and analyzing traffic in real time and reporting the current status in time provide a feasible way. In this paper we used a refined naive Bayes method, naive Bayes kernel estimator (NBKE), to identify flooding attacks and port scans from normal traffic. The mechanism of our method is based on the observation that almost all known attacks could significantly change the traffic features. Uniquely, we employ the hand-identified traffic instance as the input of the NBKE. In this paper, we illustrate the higher accuracy in detection the flooding attacks and port scan behavior by using NBKE. Our results indicate that the simplest naive Bayes (NB) estimator is able to achieve about 88.4% accuracy, while the kernel estimator can provide 96.8% accuracy. We also demonstrate that the mechanism our method based on is more reasonable.