A Hierarchical Automata Based Approach for Anomaly Detection in Smart Home Devices

Abstract

Smart home is an important application scenario of the Internet of Things. However, smart home tends to suffer from runtime failures due to its complex software/hardware, unexpected/wrong human operations, environmental changes, etc. It is hard to detect the device anomalies, since smart devices have various types, various failure types, and they also generally do not expose the inner runtime information. Current anomaly detection techniques for general programs or distributed systems do not apply for smart home devices. To help both device manufactures and users to detect the device anomalies, we propose a general framework named SmartHome-Detector. It first builds a hierarchical automata based behavior model during the testing phase, and uses it as the baseline of IoT device. At runtime, a comprehensive anomaly detection technique is proposed to identify various device anomalies in time. We implement SmartHomeDetector based on the open source smart home system HomeAssistant. Our experiments and case studies on real-world smart home devices show that SmartHomeDetector is effective. It accurately models the running behavior of IoT devices and can detect various device anomalies at runtime.