Abstract : Smart home automation, a component of the Internet of Things (IoT), enables users to manage home functions with smart sensors and actuators, providing convenience, energy efficiency, and remote monitoring. The Z-Wave protocol, widely adopted in smart homes for lighting, security, appliance control, and power management, remains vulnerable to various external attacks, highlighting the need for effective attack detection tools. The existing single-layer artificial neural network (ANN) model (i.e., ZMAD) performs well on familiar data; however, it has limitations over datasets with different distributions and advanced attack vectors.This paper introduces ZENA, a lightweight protocol-aware anomaly-based intrusion detection model for Z-Wave networks, which employs a multilayer ANN, built from scratch, to improve detection accuracy and robustness. Using a dataset with several attack classes and adversarial generated vectors, the proposed model achieves a precision rate of 95%, significantly outperforming ZMAD and state-of-the-art deep learning neural networks on same dataset (i.e., 89-93%). The results indicate substantial improvements in advanced detection and resilience against adversarial attacks, enhancing security for Z-Wave smart home systems.
Index terms : Security, Smart homes, Internet of Things, Z-Wave, intrusion detection systems, artificial neural network