Abstract:

Abstract Providing security to the Wireless Sensor Networks (WSN) is more challenging process in recent days, due to the self-organization nature and randomness of sensor nodes. For this purpose, the Intrusion Detection System (IDS) is mainly developed that supports to increase the security of network against the harmful intrusions. The conventional IDS security frameworks are highly concentrating on improving the reliability and safety of networks by using different approaches. Still, it limits with key problems of increased time consumption, more delay, and reduced efficiency, inefficient handling of large dimensional datasets, and high misclassification outputs. In order to solve these problems, the proposed work develops an intelligent IDS framework for enhancing the security of WSN by using the Cuckoo Search Greedy Optimization (CSGO) and Likelihood Support Vector Machine (LSVM) models. In this model, the most extensively used network datasets such as NSL-KDD and UNSW-NB15 are considered for validating this model. Initially, the dataset preprocessing is performed for normalizing the attributes based on the processes of irrelevant information removal, missing value prediction, and filtering. After preprocessing, the optimal number of features are selected and given to the input of CSGO algorithm, which computes the optimal fitness function for selecting the best features. Finally, the LSVM based machine learning classification technique is utilized for predicting the classified label as whether normal or abnormal. During results evaluation, the performance of the proposed security model is validated and compared by using different performance measures.

Keywords: