A Novel Anomaly Detection Method in Sensor Based Cyber-Physical Systems

Abstract

In recent times, Cyber-physical system (CPS) integrates the cyber systems and physical world for performing critical processes that are started from the development in digital electronics. The sensors deployed in CPS are commonly employed for monitoring and controlling processes that are susceptible to anomalies. For identifying and detecting anomalies, an effective anomaly detection system (ADS) is developed. But ADS faces high false alarms and miss detection rate, which led to the degraded performance in CPS applications. This study develops a novel deep learning (DL) approach for anomaly detection in sensor-based CPS using Bidirectional Long Short Term Memory with Red Deer Algorithm (BiLSTM-RDA). The presented BiLSTM-RDA model comprises preprocessing classification, and parameter tuning. Initially, the sensor data undergoes preprocessing to remove the noise present in it. Afterward, the BiLSTM based classification process takes to detect the existence of anomalies in CPS. At last, parameter tuning of the Bi-LSTM model is carried out by the use of RDA for tuning the parameters such as the number of hidden layers, batch size, epoch count, and learning rate. For assessing the experimental outcome of the BiLSTM-RDA technique, a comprehensive experimentation is performed using the data from sensor-based CPS. A detailed comparative analysis takes place to ensure the effective detection performance of the BiLSTM-RDA model and The obtained experimental results verified the superior performance on the applied data over the compared methods with the maximum an average precision of 0.989, recall of 0.984, F-score of 0.985, and accuracy of 0.983.