The paper introduces an LSTM-based Encoder-Decoder (EncDec-AD) scheme for multi-sensor anomaly detection. The method learns to reconstruct normal time-series behavior and uses the reconstruction error to detect anomalies. The authors experiment with various datasets, including power demand, space shuttle, ECG, and engine data, demonstrating that EncDec-AD can effectively detect anomalies in both predictable and unpredictable time-series. The model is robust to short and long time-series lengths, and it outperforms existing prediction-based anomaly detection models in unpredictable scenarios. The key contributions include the ability to detect anomalies from non-predictable time-series and the robustness of the LSTM encoder-decoder model.The paper introduces an LSTM-based Encoder-Decoder (EncDec-AD) scheme for multi-sensor anomaly detection. The method learns to reconstruct normal time-series behavior and uses the reconstruction error to detect anomalies. The authors experiment with various datasets, including power demand, space shuttle, ECG, and engine data, demonstrating that EncDec-AD can effectively detect anomalies in both predictable and unpredictable time-series. The model is robust to short and long time-series lengths, and it outperforms existing prediction-based anomaly detection models in unpredictable scenarios. The key contributions include the ability to detect anomalies from non-predictable time-series and the robustness of the LSTM encoder-decoder model.