Advancing Security in the Industrial Internet of Things Using Deep Progressive Neural Networks

A machine learning algorithm that can solve complex tasks by leveraging information through transfer learning while avoiding catastrophic forgetting of previously learned data. This clearly is a principal milestone to achieving human-level intelligence. The exponential growth of the Industrial Internet of Things (IIoT) has warranted an urgent need for reliable security solutions. In this paper, a novel deep progressive algorithm leveraging, both, progressive learning and deep neural networks has been proposed to achieve this. The numerous threats faced by IIoT devices are accurately classified by the proposed Deep Progressive Neural Network (DPNN), and new classes are efficiently added to the previously trained network by employing prior knowledge via lateral connections to the previously learned network. Through robust experimentation, it is shown that the DPNN model improves the efficiency and reliability of the classification of attacks and that the proposed architecture provides an accuracy of 94% compared to the 69.7% accuracy of the contemporary KNN model.