Anomaly Detection in IoT by Using Machine Learning Techniques

Abstract

The present era is marked as the era of science and technology in which IOT plays core role at a larger scale. In IOT, millions of devices are connected through wire or wireless medium that have ability to transfer data and information. With the progression in this emerging world, where millions of IOT devices are connected with each other, they have become more susceptible to different types of anomalies related to their security. The quantity of Internet of Things (loT) gadgets is developing speedily in smart homes, inculcating a lot of informational which are generally moved through wireless networks Nonetheless, different IoT devices are susceptible to risks which means they are in danger of the entirety of the terrible attacks. As the storage capacity of loT devices is limited and the processing power is low> conventional high-end solutions for loT security are not suitable. Furthermore, IoT devices are presently connected for prolonged times without human mediation. This directs that intellectual network-centered security mechanisms such as machine learning frameworks need to be established. Logistic Regression"(LR), Random Forest(RF), Artificial Neural Network (ANN),KM,„,s „„, few machine leaning models that were previously used for anomaly detection. In this study, we present a comparison of various machine learning algorithms namely K Nearest Neighbor (KNN),Decision Tree (DT), Naïve Bayes (NB), Multiplayer Perception (MLP) and Support Vector Machines(SVM) that can be used to detect malicious or abnormal data in loT and provide the best algorithm. Further, we proposed ensemble model of above mentioned 5 machine learning (ML) algorithms that can be utilized to rapidly and adequately identify the IoT attack„hDD’99 cup dataset was used for experimentation. The outcomes are then compared with previous anomaly detection studies and provide improved results. Our experiment results highlight that Ensemble model is the best model, among all other machine learning models that are used in this study .