Enhancing the Performance of an Intrusion Detection System Using Spider Monkey Optimization in IoT

Abstract

The Internet of Things (IoT) has led to an era with the development of communication between smart devices used in various fields. Due to insufficient measures in the infrastructure of smart devices, it is prone to various attacks, which are launched by intruders during data transmission through the internet. Therefore, providing security measures for IoT systems is considered the highest priority. The present work ensures an Intrusion Detection System (IDS), which monitors malicious activities and helps in the successful functioning of IoT networks. Therefore, Spider Monkey Optimization with Random Forest (SMO-RF) algorithm is proposed to detect attacks in IoT environments. The NSL-KDD dataset is used where the input data is pre-processed and then classification is done using Random Forest (RF). The SMO will consider the features from where the decision for splitting or combining of the data taken by the female leader based on the 80-20 rule at the Global Leader Phase (GLP). Calculating the feature or variable importance with a Random Forest model shows which and all the features of the data are the most helpful for Classification is used. Thus, proposed SMO selects the best highest feature importance value for the classification of attacks using RF that overcomes the problem of over fitting. The results obtained from the proposed SMO-RF show the accuracy of 99.98 % better when compared with the existing SVM-parameter optimization and PSO-multi SVM techniques of 99.8% and 98 % respectively