Swarm Intelligence Based Feature Selection for Intrusion and Detection System in Cloud Infrastructure

Network intrusion and cyber attacks are the most severe concern for Cloud computing service providers. The vulnerability of attacks is on a hike that manual or simple rule-based detection of cyber-attacks is not robust. In order to tackle cyber attacks in a reliable manner, an automated Intrusion Detection system equipped with a swarm intelligence (SI) based machine learning model (ML) is essential to deploy at entry points of the network. Nowadays, the application of SI with ML is used in various research areas. For an efficient IDS, choosing relevant features from the noisy data is an open question. In this regard, this paper proposes a method that utilizes the Whale Pearson hybrid feature selection wrapper for reducing the irrelevancy in the IDS model. Whale Pearson hybrid wrapper is an improved version of the binary Whale optimization Algorithm (WOA). The WOA is a type of SI algorithm which is inspired by the behavior of humpback whales. The proposed method has chosen 8 out of 42 features from the Hackereath Network attack prediction data-set, which are sufficient for building an efficient Intrusion detection model. The model trained with the eight features produces an accuracy of 80%, which is 8% greater than the accuracy produced by the original data-set with the KNN algorithm on ten-fold cross-validation.