An Adaptive Intrusion Detection System in Industrial Internet of Things(IIoT) using Deep Learning

The Industrial-Internet-of-Things (IIoT) is a product of the extensive use of the Internet-of-Things(IoT) in vital industries including manufacturing and industrial production. To improve industrial and manufacturing processes, the IIoT integrates sensors, actuators, and smart tools that can interact with each other. There are many advantages to IIoT for service providers and customers alike, but privacy and security are still major concerns. Cyberattacks in such a network have been reduced with the use of an intrusion detection system (IDS). Nevertheless, several current IIoT intrusion detection systems (IDS) suffer from issues such as an incomplete list of the network's attack kinds, an excessive number of features, models constructed using outdated datasets, and an absence of attention to the issue of imbalanced datasets. Our proposed intelligent recognition system can spot cyberattacks in IIoT-networks, which will help with the difficulties. Singular value decomposition (SVD) is employed by the suggested model to decrease data characteristics and enhance detection outcomes. If we want to keep our classifications from being biased due to over-fitting or under-fitting, we employ the SMOTE method. Data has been classified using a number of deep learning and machine learning techniques for both binary and multi-class purposes. We test the suggested intelligent model on the ToN_ IoT dataset to see how well it performs. With the suggested method, we were able to achieve a 99.98% accuracy rate and a lowered error rate of 0.016% for multi-class classification, and a 0.001 % reduction in the error rate for binary classification.