KPI (Key Performance Indicator) Sensor is a device or system designed to monitor and measure specific metrics or indicators that are crucial for assessing the performance of a process, system, or organization.

Posts

State-Aware Anomaly Detection for Massive Sensor Data in Internet of Things

With the escalating prevalence of Internet of Things (IoTs) in critical infrastructure, the requirement for efficient and effective anomaly detection solution becomes increasingly important. Unfortunately, most prior research works have largely overlooked to adapt detection criteria for different operational states, thereby rendering them inadequate when confronted with diverse and complex work states of IoTs. In this study, we address the challenges of IoT anomaly detection across various work states by introducing a novel model called Hybrid State Encoder-Decoder (HSED). HSED employs a two-step approach, beginning with identification and construction of a hybrid state for Key Performance Indicator (KPI) sensors based on their state attributes, followed by the detection of abnormal or failure events utilizing high-dimensional sensor data. Through the evaluation on real-world datasets, we demonstrate the superiority of HSED over state-of-the-art anomaly detection models. HSED can significantly enhance the efficiency, adaptability and reliability of IoTs and avoid potential risks of economic losses by IoT failures.