A Binary Bee Foraging Algorithm-Based Feature Selection Approach for IoT Intrusion Detection

Intrusion detection system (IDS) is a security solution to detect the anomalies of the Internet of Thing (IoT) network and raise alerts. Feature selection is one of the essential preprocess steps to build an efficient intelligent intrusion detection system (IDS), which is helpful to reduce the number of features (NF) that are used to classify attacks. The meta-heuristic algorithm-based approaches has been widely used to solve feature selection problems in the past few years. However, as the data dimension increases, the dramatically expanded search space challenges the optimization algorithms, including the recently proposed bee foraging algorithm (BFA). In this article, a binary variant of BFA (BBFA) is presented to binarize BFA. BBFA utilizes three diverse mechanisms to update the position of three different types of bees, which helps to promote the evolutionary efficiency. First, a Sigmoidal transfer function is utilized to update the positions of scout bees. Second, a V-shaped transfer function is applied to binarize the positions of forager bees. Third, a binary update strategy is proposed to make the onlooker bees efficiently search the space. The performance is evaluated using 22 well-known public machine learning data sets to compare the proposed approach with several similar variants of meta-heuristic algorithms of the literature in terms of fitness function values, classification accuracy, number of selected features, computational cost, and convergence properties. The results show that the proposed approach is able to find the subsets with high classification accuracy and few NF while also requiring less computational load.