The physical process with the digital computing channel and electronic computing is integrated by the Cyber-Physical Systems (CPS). The abnormality and failures are the two main sources that affect the performance of ...
详细信息
The physical process with the digital computing channel and electronic computing is integrated by the Cyber-Physical Systems (CPS). The abnormality and failures are the two main sources that affect the performance of the CPS. In the CPS, the research on security analysis and fault diagnosis has attracted lots of interest from the researcher. However, the existing model does not find the difference between the fault and attacks, the existing approaches need adequate development for identifying the difference between the fault and attack in the Internet of Things (IoT)-CPS. In this research, a deep learning-based approach is developed to detect the attacks and faults in the IoT-CPS for enhancing the security of the network. The IoT-based data is garnered from a standard online source. After collecting the data, the extraction of the deep features is performed using the Conditional Variational Autoencoder (CVA). The attained deep attributes are further taken for the weighted feature fusion process in which the required weights are chosen in an optimal manner using the Enhanced Egret Swarm Optimization (EESO) algorithm. The obtained weighted fused features are inputted into the Dilated Convolutional Neural Network and bidirectional long short-term memory with multi-scale dense attention (DCNN-Bi-LSTM-MSDA). The classification outcomes are obtained from the DCNN-Bi-LSTM-MSDA module. Throughout the result analysis, the accuracy and NPV rate of the designed model is 94.16% and 99.38%. The validation of the fault and attack classification offered by the implemented deep learning-oriented failure and abnormality classification scheme in IoT-based CPS is done against several traditional models.
暂无评论