LightGBM-RF: A Hybrid Model for Anomaly Detection in Smart Building

Abstract

Smart building uses sophisticated and integrated building technology and allows numerous IoT systems to interact as well as provide convenience to its users. Unfortunately, smart buildings have become a point of attraction for cybercriminals. Due to the fact that the majority of these IoT devices lack the memory and computing power required for robust security operations, they are inherently vulnerable. IoT devices are consequently vulnerable to various attacks. Therefore, a single attack on network systems or devices can cause serious harm to the security of data as well as privacy in a smart building. This paper presents LightGBM-RF, a machine learning model that accurately detects anomalies in a smart building by utilizing a combination of Light Gradient Boosting Machine and Random Forest algorithms. The model detects anomalies with an accuracy of 99.19%, thereby providing an effective scheme for detecting different attack families, and the potential to significantly improve security in smart buildings.