Anomaly Detection and Mitigation of Security Issues in IoT network

Abstract

This research work primarily focuses on optimizing the length of independent newlinefeature set, served as input to the Machine Learning algorithm for efficiently newlinedetecting anomalies in real time in IoT networks. Along with this, the researchers newlineaimed to improve the anomaly detection accuracy and other standard metrics, both newlinein overall terms and for some of the individual classes of attacks. In essence, the newlineresearchers were motivated to efficiently and effectively detect the anomalies or newlineattacks directed at IoT network and thereby mitigate any security issues arising out newlineof these attacks. newlineIn the first study, an intelligently initialised hybrid binary PSO-GWO method newlinebased on random forests is proposed to increase the overall accuracy of anomaly newlineidentification. The concept or idea of the leaders relative weights in GWO newlinewas applied to update the positions of the particles in PSO. In addition, a new newlinefitness function was proposed that integrates the critical performance metrics for newlinedetermining the success of categorisation. For commensurate comparison with newlineother comparable research, a new performance metric was proposed in the paper. newlineThe findings indicate a general improvement in network anomaly detection. newlineIn the second study, investigation of comparative efficiency of SVM Kernels and newlineparameters was taken up and the best parameters were applied to the proposed newlinemodified binary Gray wolf algorithm (RbGWO) for fitness measurement. The newlinemodified binary Gray wolf algorithm (RbGWO) was implemented to pick only a newlinesubset of the features in the original dataset as opposed to the entire set of features. newlineIn this work, the positions of non-leader wolves was updated by changing the newlineix newlinenumber of leaders.... newline