Perspective on efficiency enhancements in processing streaming data in industrial IoT networks

Abstract

Both data compression and anomaly detection are very deeply studied areas for the last decades and gain significance for the Internet of Things (IoT), especially industrial IoT (IIoT). Due to the advantages like fewer latency and security aspects, edge computing is often preferred to cloud solutions. While the amount of data as well as the demand for edge data processing increases, resources like bandwidth, computational performance, memory and, in case of Wireless Sensor Networks (WSN), also energy are still limited. This leads primarily to a trade-off between maximum data reduction, information extraction and minimal computational effort. Often, both data compression and anomaly detection are required. This paper demonstrates additional benefits if already one is implemented. Although in many cases the algorithms for both are based on the same models, there are almost no studies on their combined use. In this work, a perspective on the efficiency of combined model usage with only different interpretations for anomaly detection and data compression is proposed. Concrete examples for selected models and the detection of different kinds of anomalies are given. Finally, an outlook on the planned future work is given.