Simulation and Modeling for Anomaly Detection in IoT Network Using Machine Learning

Today we are living in an era where everything is changing to be smart, whether it be a smart home, smart industries, smart irrigation, or a smart meter, where the word smart refers to the involvement of the Internet of Things (IoT). The increased use of IoT infrastructure in these fields has led to the failure of the nodes, increase in threats, attacks, abnormalities, and spying, which is the primary concern and an important domain of an IoT. The main objective of this paper is to use a supervised learning model to predict anomalies in the historical data which can later be incorporated into real-world scenarios to block the upcoming anomalies and attacks. This paper predicts the anomalies on the 350 K data set using the Machine Learning models and compares its performance based on the state of arts. In this paper, two different approaches are used based on the analysis done on the dataset. The classification algorithms were applied to the whole dataset in the first, and then the same classification algorithms were applied after excluding the data points having binary values (0 and 1) in the feature "value" and have achieved an average of 99.4% accuracy for the first case and 99.99% accuracy for the later.