An efficient feature selection based Bayesian and rough set approach for intrusion detection

The exponential growth of network size leads to increase attacks and intrusions. Detection of these attacks from the network has turned into a noteworthy issue of security. An intrusion detection system is an important approach to achieves high detection rate. A high dimensional dataset increase complexities of detection systems. In this paper, we have designed a novel intelligent system that comprises the feature selection with a hybrid approach of the Rough set theory and the Bayes theorem. The proposed feature selection computed core features and ranked them based on estimated probability. In a decision system, an object may belong to a single or multiple decision, and a feature contains a set of objects that occurrences compute an estimated probability. The rough set theory is being applied to classify information into lower and upper approximations. Uncertain information is distinguished using rough set approximations and solved by the Bayes theorem. In this research work, it has also been highlighted the quantitative realism of recently generated dataset and compared to publicly available datasets. This approach reduces false alarm rate, computational complexity, training complexity and increases detection rate. Comparisons with relevant classifiers are also tabled that show proposed method performs better than existing classifiers.