Poster Abstract: Explainable Sensor Data-Driven Anomaly Detection in Internet of Things Systems

Deep learning or black-box models are widely used for anomaly detection in Internet of Things (IoT) data streams. We propose a technique to explain the output of a deep learning model used to detect anomalies in an IoT based industrial process. The proposed technique employs dual surrogate models to deliver black box model explanation. We have also developed an interactive dashboard to give further insights into the detected anomaly. The dashboard integrates our proposed deep learning explanation technique with historical logs to explain the detected anomaly for personas with different backgrounds.