Developing Anomaly Detection on IoT Devices Using Machine Learning (ML)

Abstract

Many experts have explored the risks posed by Internet of Things (IoT) devices to large corporationsand smart cities. Because of the rapid acceptance of IoT andthe nature of these devices, their inherent mobility, and the limits imposed by standardization, sophisticated systems capable of detecting suspicious movement on IoT devices connected to a network are necessary. As the number of Internet of Things devices connected to the Internet increased, so did the capacity of Internet traffic. As a result of this change, typical methods and traditional data processing approaches for detecting attacks are no longer valid and should be avoided.

Because of the increased volume of network data, detecting assaults in the Internet of Things (IoT) and identifying malicious activity in its early stages is a particularly tough problem to tackle. This article proposes and offers evidence for an approach for identifying malicious network traffic. For identifying malicious network traffic, the framework employs three commonly used classification-based approaches. The Support Vector Machine (SVM), Random Forest (RF), and logistic regression (LR) algorithms all execute with 100% accuracy. The dataset Botnet-IoT was employed in the model creation used in this study framework, and the results in terms of training, specificity, and accuracy were compared.

Keyword: IoT, LR, SVM, RF, Botnet-IoT