Attack Detection in Industrial IoT via Deep Hybrid Model on Solving Class Imbalance Problem

The Industrial Internet of Things (IIoT) emerged as an acceptable data processing architecture as several sectors have gone digital. Nevertheless, there are still difficult problems with IIoT, such as the reliability of 3rd parties, data authenticity, effectiveness 'as well as system reliability. This paper proposes a deep hybrid model based industrial IoT Attack Detection (DHM-IIoTAD) by following four steps like preprocessing, feature extraction, attack detection, & mitigation. Moreover, the pre processing is the initial step to solve the class imbalance problem. Then, the features including Correlation based feature, and higher order statistical based features are extracted during the feature extraction phase. Further, the deep hybrid model combining the models like Deep maxout & Bi- LSTM model is utilized during the attack detection stage, which determines the final detection outcome with Improved Score level fusion. After the detection process, the mitigation is performed via entropy-based mitigation process. Finally, the performance of DHM-IIoTAD scheme is compared over extant models based on certain metrics including FNR, sensitivity, NPV, accuracy, FPR, precision, specificity, MCC, and F-measure, respectively for WUSTL-IIOT-2018 Dataset and IoT Intrusion Dataset.