A Linear Nearest Neighbour Lasso Step Based Honey Badger Algorithm for Robust Network Intrusion Detection System

Nowadays, one of the biggest issues facing network-based intrusion detection systems are handling massive volumes of network data (IDS). These data typically have a large number of duplicated or unnecessary attributes. Using feature selection techniques, pertinent features must be extracted from the original data in order to increase the effectiveness of IDS. Feature selection is a crucial component of intrusion detection and helps to enhance intrusion detection performance. For feature selection of network intrusion detection, a Honey Badger method based on Linear Nearest Neighbour Lasso Step (LNNLS-HB) is presented to address the issue of low efficiency and high false positive rate in IDS caused by increasing high-dimensional data. The LNNLS-HB algorithm's fitness evaluation function incorporates the quantity of characteristics used and the precision of the classification. Finally, the Convolutional Neural Network (CNN) classification is performed optimal solution, the updated HB location is subjected to the linear closest neighbour lasso step optimisation. To assess and compare the performance of the proposed methodology with other widely used feature selection techniques, the intrusion detection benchmark dataset (CIC-IDS2017) is used. Empirical results demonstrate that the proposed LNNLS-HB achieves a better accuracy of 99.52% when compared with the existing methods.