Risk analysis in cyber situation awareness using Bayesian approach

Abstract

The unpredictable cyber attackers and threats have to be detected in order to determine the outcome of risk in a network environment. This work develops a Bayesian network classifier to analyse the network traffic in a cyber situation. It is a tool that aids reasoning under uncertainty to determine certainty. It further analyze the level of risk using a modified risk matrix criteria. The classifier developed was experimented with various records extracted from the KDD Cup'99 dataset with 490,021 records. The evaluations showed that the Bayesian Network classifier is a suitable model which resulted in same performance level for classifying the Denial of Service (DoS) attacks with Association Rule Mining while as well as Genetic Algorithm, the Bayesian Network classifier performed better in classifying probe and User to Root (U2R) attacks and classified DoS equally. The result of the classification showed that Bayesian network classifier is a classification model that thrives well in network security. Also, the level of risk analysed from the adapted risk matrix showed that DoS attack has the most frequent occurrence and falls in the generally unacceptable risk zone.