An Efficient Anomaly Detection Framework for Electromagnetic Streaming Data

Degang Sun, Yulan Hu, Zhixin Shi, Guokun Xu, Wei Zhou

Abstract: The electromagnetic signal is a kind of communication signal whose intensity can reflect the electromagnetic condition of the current space. The continuous electromagnetic signal data can be seen as a kind of time-series flow data. During actual monitoring, the use of wireless devices can affect the electromagnetic space distribution and alter the signal strength at the corresponding frequency point. By detecting the changes of signal strength, the use of wireless devices can be effectively found. Based on the isolation forest algorithm and the Storm streaming computing platform, an efficient framework for wireless devices usage detection is proposed. The experiments results based on the real-world electromagnetic streaming data demonstrate that the framework can accurately and efficiently detect abnormal signals.