PEAR: Positional-encoded Asynchronous Autoregression for satellite anomaly detection

This paper proposes a Positional-Encoded Asynchronous AutoRegression (PEAR) method for satellite anomaly detection. We empirically observe that a single classification model can hardly detect unknown anomalous situations and neglect the Markov nature of temporal satellite data. To address this, we adopt an autoregressive model to deal with the prediction of unknown anomaly for satellite data. We further propose a non-uniform temporal encoding method for asynchronous data and a median filtering method for more accurate detection. To reduce the effect of outliers, we employ an adaptive threshold selection method to achieve a more robust classification boundary. We test the proposed method on the time series prediction models including LSTM and transformers and we further employ the positional encoding strategy to improve the modeling capabilities of the Transformer model on high-frequency information. Experiments on real satellite data demonstrate that the proposed PEAR method outperforms the baseline method by 55.79%.