Anomaly detection based self-healing mechanism using dynamic diffusion spatial-temporal graph convolutional network in industrial IoT

The Industrial Internet of Things (IIoT) has been viewed by the public as the key component of Industry intelligent digital factories. To guarantee the security and resilience of the IIoT, which has many susceptible IIoT devices, effective anomaly detection is essential. In this paper, Anomaly Detection Based Self-Healing Mechanism Using Dynamic Diffusion Spatial-temporal Graph Convolutional Network in Industrial IoT (AD-SHM-DDSGCN-IIoT) is proposed. At first, Input data is collected from WUSTL-IIOT-2021 Dataset. Then, using Sparse Regression Unscented Kalman Filter (SRUKF) which is used to reduce the noise and standardize from the collected data. The pre-processed data are given into Dynamic Diffusion Spatial-temporal Graph Convolutional Network (DDSGCN) for detecting the anomaly as Normal Traffic, Total Attack Traffic, DoS Traffic, Reconnaissance Traffic, Command Injection Traffic, and Backdoor Traffic. To ensure precise anomaly detection from Industrial IoT, DDSGCN generally does not express any adaptation of optimization strategies for figuring out the ideal parameters. Starfish Optimization Algorithm (SFOA) is proposed in this work to optimize the weight parameter of DDSGCN classifier, which precisely classifies the anomaly from Industrial IoT. The proposed AD-SHM-DDSGCN-IIoT is implemented and analyzed under performance metrics, such asAccuracy, Recall, Precision, Fault Recovery Time, Network Throughput and ROC. The performance of the AD-SHM-DDSGCN-IIoT approach attains24.18 %, 31.66 % and 18.46 % higher accuracy and 18.16 %, 25.49 % and 30.68 % lower fault recovery time with existing methods respectively.