Abstract
Malicious and anomaly traffic prediction in the Internet of Things (IoT) is substantial for establishing IoT in optical network security. The researchers propose various learning approaches to develop security to handle the malicious traffic flows in optical network. This research concentrates on modelling an efficient feature selection approach using correlation mapping. This mapping process is achieved by enhancing the conventional wrapper method to filter out the unnecessary features and select the most influencing for the different classification processes. Then, the entropy-based information analysis is performed to validate the chosen features of optical data to predict malicious traffic over the network. The evaluation is done with an online available Botnet-IoT dataset, and the comparison is made with various other ML approaches. The experimental outcomes show that the anticipated model improves prediction outcomes more effectively than different processes with better trade-offs.