A Review of the Application of Data Science Methods in Communication Signal Anomaly Detection

With the large-scale deployment of 5G and industrial Internet of Things devices, the electromagnetic environment has become complex and diverse, and communication signal anomalies caused by jamming interference, spoofing interference, and equipment faults have become increasingly prominent. Traditional threshold and time-frequency analysis methods depend on manual experience, and their performance is limited in practical scenarios with unknown interference, low signal-to-noise ratios, and scarce samples. In recent years, data science technologies represented by few-shot learning, open-set recognition, lightweight deep models, and improved unsupervised frameworks have become research hotspots in communication signal anomaly detection. This study reviews recent domestic and international research results from three dimensions: few-shot optimization under supervised paradigms, architectural improvement of unsupervised reconstruction models, and open semi-supervised models. It compares the advantages and existing shortcomings of different technical routes, analyzes practical challenges related to samples, signal-to-noise ratios, and deployment in real electromagnetic environments, and looks ahead to future development directions.