Anomaly Detection in IoT Sensor Networks using Machine Learning

Anomaly Detection is a promising new approach for quality control in wireless networks and telecommunication networks. New modern network architecture enables intensive computing and communication at the edge of the network. It can thus support large scale IoT applications. The proper and timely detection of attacks in the Internet of Things (IoT) infrastructure is very important. With the rise in IoT technologies in various application areas these days, the attacks and threats are also on the verge of exploiting this. Denial of Service, Data Type Probing, Malicious Control, Malicious Operation, Scanning, Spying and Wrong Setup are the different attacks and anomalies which can cause a communication system failure. In the past decades an enormous rise in the internet applications was seen, which has led to concreting the need of security of the information networks by many folds. As the network operations continuously changes in different time frames, an anomaly detection system should be capable of adapting dynamically to the situations. The objective of this paper is to identify various attacks on IoT sensor networks and the development of a generalized anomaly detection model using machine learning techniques to check if an IoT sensor network is anomalous or not. Kaggle and NSL-KDD datasets are used for this study. Model is trained based on the Levenberg – Marquardt optimization using TRAINLM training function and LEARNGDM adaptation learning function. High performance was exhibited by the model. Menu based environment is also created in Jupyter in this paper.