Feature Selection by Hybrid Artificial Bee Colony Algorithm for Intrusion Detection

The fast increase in number of different hardware being attached to the internet of things (IoT) mesh also leverages the importance of solving the security issues for establishing the secure environment data transfer over the network. As novel vulnerabilities are being discovered daily, the scientists and network security specialists must devise new models to detect and prevent the intrusions and keep the network safe. One way to address this challenge is to apply the artificial intelligence to develop modern intrusion detection systems (IDS). These cutting-edge systems are able to adjust and counter the novel threats as they come to light, allowing reaction in the real time and altering their bearing with respect to the previous experience. However, the challenge to classify the ongoing traffic is getting more complex due to the immense data volume that is produced by the network systems and leveraged computation requirements. Consequently, the feature selection must be employed as a mean to decrease data complexity through the process of removal of the least relevant information for the current classifying job, increasing the accuracy of the model on the way. This research suggests using a hybrid variant of the commonly used artificial bee colony (ABC) metaheuristics to tackle the feature selection, aiming to increase the accuracy of the extreme learning machine classifier. The novel model has been validated on two famous network security datasets (UNSW-NB15 and CICIDS-2017) to demonstrate the improvement in the performances, and the simulation outcomes were compared to the results obtained by other contemporary methods that have been employed for the same task and under similar circumstances.