Machine Learning Approach to Anomaly Detection Attacks Classification in IoT Devices

Abstract

Internet of Things (IoT) devices have a dark side could be used against the users or threaten the user by hackers and intruders. In addition, IoT devices have some security issues because the devices are basically connected to the internet and are more likely to get mishandled by hackers using anomaly attacks. In this paper, we proposed the application machine algorithms to detect anomaly attacks in IoT devices. The selected algorithms include are the Support Vector Machine (SVM) and Random Forest (RF). The SVM and RF are powerful supervised learning method that was utilized for both detection and feature selection. A standard anomaly dataset called the NSL-KDD dataset was used for the experimentation in arff format. The results shows an accuracy of approximately 99.9% and 97.9% with RF and SVM respectively, while a false positive rate of 0.1% was achieved in all scenarios for classification of anomaly attacks in IoT devices. This shows that the proposed method RF has higher accuracy than previous literatures, which is very promising. The RF and SVM posted a very encouraging recall and precision as well.