An Efficient Method to Decide the Malicious Traffic

Abstract

To address the high rate of false alarms, this article proposed a voting-based method to efficiently predict intrusions in real time. To carry out this study, an intrusion detection dataset from UNSW was downloaded and preprocessed before being used. Given the number of features at hand and the large size of the dataset, performance was poor while accuracy was low. This low prediction accuracy led to the generation of false alerts, consequently, legitimate alerts used to pass without an action assuming them as false. To deal with large size and false alarms, the proposed voting-based feature reduction approach proved to be highly beneficial in reducing the dataset size by selecting only the features secured majority votes. Outcome collected prior to and following the application of the proposed model were compared. The findings reveal that the proposed approach required less time to predict, at the same time predicted accuracy was higher. The proposed approach will be extremely effective at detecting intrusions in real-time environments and mitigating the cyber-attacks.