Detecting Anomalies in IOT Devices through Machine Learning Techniques

Abstract

The rapid increase in Internet of Things devices has sparked fears of vulnerability in various areas, including anomaly detection. This paper introduces a novel use of machine learning algorithms, Support Vector Machines and Random Forests, as well as ensemble methods such as stacking and voting classifiers for anomaly detection on the Internet of Things. Based on the NSL-KDD dataset, the experiment demonstrates the efficacy of RF and stacking classification methods, achieving high accuracy with fewer false positives than current literature. Both ensemble methods, Voting Classifier RF + AB and Stacking Classifier RF + MLP with LightGBM, achieve exceptional performance, recall and precision, proving suitable for identifying and managing anomalous behaviours in various systems. Moreover, the project includes integrating user evaluation through a Flask framework front-end with user authentication, a critical component of IoT anomaly detection's practical implementation. This paper demonstrates the potential of ML approaches in improving the security and endurance of the Internet of Things by efficiently identifying and managing variations in multitudes of Internet of Things working environments.