Abstract: For satellite telemetry data, imbalance of distribution and lack of anomaly tags are main problems. Therefore, usual anomaly detection method is not effective in this respect. An anomaly detection method based on timing generation countermeasure network is proposed. Preprocessed to eliminate the noise and outliers, the satellite telemetry data is used to train the generator and discriminator, so that the time dependence of historical data can be learned. Finally, the adaptive threshold method is used to determine whether the residual of generator is abnormal. The experimental results show that the anomaly detection method is effective.

Key words: satellite telemetry data, anomaly detection, generative adversarial network, adaptive threshold